OpenSIPS MCP Server
OfficialGenerates a complete observability stack including Grafana dashboards for monitoring OpenSIPS performance, with provisioning and a docker-compose setup.
Provides database management capabilities for OpenSIPS tables, enabling CRUD operations on subscribers, dispatchers, dynamic routing, permissions, call center, load balancer, dialplan, and accounting data stored in MySQL.
Provides database management capabilities for OpenSIPS tables, enabling CRUD operations on subscribers, dispatchers, dynamic routing, permissions, call center, load balancer, dialplan, and accounting data stored in PostgreSQL.
Generates a complete observability stack including Prometheus metrics collection and configuration for monitoring OpenSIPS, with provisioning and a docker-compose setup.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@OpenSIPS MCP Server list available configuration scenarios"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
OpenSIPS MCP Server
An AI-powered and shell-callable control plane for OpenSIPS. Connect any MCP-compatible AI client (Claude Desktop, Cursor, Cline, Continue, Cody) for natural-language operations — or use the same primitives via the opensips-mcp CLI for shell scripts, CI pipelines, and air-gapped operators.
Presented at OpenSIPS Summit 2026, Bucharest, April 28-30.
# AI client mode — connect any MCP-aware client
opensips-mcp serve
# CLI mode — no LLM required
opensips-mcp cfg list-scenarios
opensips-mcp cfg generate residential_pbx --param db_url=mysql://... --param domain=pbx.example.com
opensips-mcp cfg lint /etc/opensips/opensips.cfg --fail-on warning
opensips-mcp snapshot capture /var/lib/snapshots --label pre-deployFeatures
Category | Count | Description |
Tools | 236 | MI, DB CRUD, cfg engine, migration, linting, tuning, diagnostics, security, B2BUA, presence, events, snapshots, benchmarks, RBAC, observability stack generator (Grafana + Prometheus + provisioning + monitoring docker-compose) |
Scenarios | 25 | From class-4 SBC and residential PBX to IMS Cx with HSS interop, STIR/SHAKEN, SIPREC, MSRP, RADIUS auth, qrouting, RTT accessibility |
Prompts | 22 | Config building, migration, troubleshoot, security audit, education |
Operator-facing docs | 15 | SIP primer, error-code decoder, annotated cfg tour, ngrep/sngrep, end-to-end PBX tutorial, HA failover, NAT decision tree, TLS-from-scratch, post-deploy smoke test, IMS Cx with open5gs, compliance positioning, observability stack, Kamailio comparison + migration, limitations, more |
Compliance frameworks addressed | 8 | HIPAA, GDPR/CCPA, SOC 2, PCI DSS, FCC § 13.114 (E911), STIR/SHAKEN, RFC 4103 RTT, NENA i3 |
Tests | 751 | Drift-prevention regression markers for every audit finding |
Skills | 6 | Claude Code slash commands |
Subagents | 6 | Claude Code orchestration roles |
Why this over X?
If you're considering... | This project gives you |
Hand-written m4 templates | A typed scenario library + auto-discovered knobs + a linter that catches what |
| Same primitives + 25 scenario templates + the migration engine + drift snapshots + 15 operator-facing docs |
Kamailio ( | An honest comparison (docs/learn/kamailio-vs-opensips.md) and a migration guide (docs/learn/migrate-from-kamailio.md) |
Twilio / Telnyx / Bandwidth | Self-hosted control over your SIP — at the cost of operating it yourself |
DIY without an LLM | The CLI is the same code path; nothing requires AI |
DIY without a CLI | Same primitives via MCP for any AI client |
Skeptical or looking for what this is NOT good for? Read docs/learn/limitations.md before adopting.
Highlights
Full MI Coverage -- Execute any Management Interface command, list commands, gather statistics, and run health checks.
Database Management -- CRUD operations for subscribers, dispatchers, dynamic routing, permissions, call center, load balancer, dialplan, and accounting tables.
Configuration Engine -- Generate complete OpenSIPS configs from 8 scenario templates plus the m4/local.cfg override pattern, with a Kamailio-compatible
WITH_*feature-flag composer that materialises a flat config (no#!ifdefnesting hell).Version-Aware Migration -- Declarative per-hop rule chain across 2.4 → 3.0 → 3.1 → 3.2 → 3.3 → 3.4 → 3.5 → 3.6 (7 fully-populated hops). Rewrites behaviour-preserving changes (MI command renames, AVP prefix drop, global param renames) and flags every manual step / silent-gotcha the engineer must handle by hand. Includes the painful 3.4 → 3.5
load_balancer_* → lb_*MI rename that silently breaks monitoring in 3.6. An eighth hop 3.6 → 4.0 is scaffolded with the MI renames anticipated in the 4.0 beta but emits an "advisory" warning until a stable 4.0 lands. (OpenSIPS skipped 3.7 entirely — the chain reflects the actual release line.)Semantic Linter -- 20+ rules checking module dependency graph, orphaned
modparam, deprecated parameters, missingmf_process_maxfwd_header, anti-flood absence, NAT/rtpengine gaps, and production readiness.Capacity Tuner (PGTune for SIP) -- Describe your workload (role + cores + RAM + CPS + NAT + TLS + HA), get a paste-ready override snippet with a
# why:comment on every line.Module Catalog + Explain -- Machine-readable module metadata powers a
kubectl explain-style lookup for directives, modparams, script functions, pseudo-variables, and MI commands (with rename history).Diagnostics -- Process lists, memory stats, TCP connections, uptime, tracing, and comprehensive health checks.
Security -- RBAC with readonly/admin roles, input validation, HA1 digest generation, config audit, Pike anti-flood management, and structured audit logging.
Version Support -- 2.4 → 3.6 fully rule-backed; one anticipatory 3.6 → 4.0 hop with master-branch MI renames (advisory until 4.0 stable). Runtime version strategies for 3.4 and 3.6 (production-ready) plus a 4.0 strategy flagged as experimental.
Docker Ready -- 7 pre-built Docker scenarios and a full-stack compose for instant deployment.
Related MCP server: oitvoip-mcp-server
Quick Start
pip install
# Install from source
pip install -e .
# For MySQL support
pip install -e ".[mysql]"
# For PostgreSQL support
pip install -e ".[postgres]"Configure
cp .env.example .env
# Edit .env with your OpenSIPS MI URL and database connectionKey environment variables:
Variable | Default | Description |
|
| OpenSIPS MI HTTP endpoint |
|
| Database connection string |
|
| Target OpenSIPS runtime version (3.4 or 3.6) |
|
| MCP transport (stdio, sse, streamable-http) |
|
| RBAC role (readonly, admin). Defaults to least-privileged; set to |
|
| Logging level |
|
| Bind address for network transports |
|
| Listen port for network transports |
| (empty) | Optional API key for authentication |
|
| Path to opensips binary for config validation |
|
| Force read-only mode regardless of role |
Run
# stdio transport (for Claude Desktop, Cursor, etc.)
opensips-mcp
# SSE transport (for web-based clients)
opensips-mcp --transport sse --host 0.0.0.0 --port 8080
# Streamable HTTP transport
opensips-mcp --transport streamable-http --host 0.0.0.0 --port 8080MCP Client Integration
~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or
%APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"opensips": {
"command": "opensips-mcp",
"env": {
"OPENSIPS_MCP_MI_URL": "http://your-opensips:8888/mi",
"OPENSIPS_MCP_DB_URL": "mysql+asyncmy://opensips:opensipsrw@localhost/opensips"
}
}
}
}claude mcp add opensips -- opensips-mcp --transport stdioOr edit .claude/settings.json:
{
"mcpServers": {
"opensips": { "command": "opensips-mcp", "args": ["--transport", "stdio"] }
}
}The project ships skills and subagents in .claude/ — Claude Code picks them
up automatically when run inside the repo.
~/.cursor/mcp.json:
{
"mcpServers": {
"opensips": { "command": "opensips-mcp", "args": ["--transport", "stdio"] }
}
}~/.continue/config.json:
{
"experimental": {
"modelContextProtocolServers": [
{
"transport": {
"type": "stdio",
"command": "opensips-mcp",
"args": ["--transport", "stdio"]
}
}
]
}
}opensips-mcp --transport sse --host 0.0.0.0 --port 8700Point any MCP-speaking client at http://<host>:8700/sse. Put an auth proxy
in front in production — see docs/security.md and
docker/nginx/ for a working config.
Docker Quick Start
Start OpenSIPS, MySQL, and the MCP server with a single command:
docker compose -f docker/docker-compose.yml up -dThis brings up:
opensips -- OpenSIPS 3.6 with MI HTTP on port 8888
mysql -- MySQL 8.0 with the OpenSIPS schema
opensips-mcp -- MCP server on port 8080 (streamable-http transport)
See docs/deployment.md for production deployment guidance.
MCP Tools
MI Commands
mi_execute-- Execute any MI commandmi_list_commands-- List available MI commandsget_statistics-- Retrieve runtime statisticshealth_check-- Comprehensive health check
User Location
ul_show_contacts-- Show registered contacts for an AORul_dump-- Dump all registered usersul_add_contact-- Add a contact bindingul_remove_contact-- Remove a contact binding
Dialog Management
dlg_list-- List active dialogsdlg_end-- Terminate a dialogdlg_get_profiles-- Get dialog profile sizesdlg_push_var-- Set a dialog variable
Subscriber Management
subscriber_list-- List subscribers (with pagination)subscriber_get-- Get subscriber by IDsubscriber_create-- Create a new subscribersubscriber_update-- Update subscriber fieldssubscriber_delete-- Delete a subscriber
Dispatcher (Load Balancing)
dispatcher_list_db-- List dispatcher destinationsdispatcher_add-- Add a destinationdispatcher_update-- Update a destinationdispatcher_remove-- Remove a destination
Dynamic Routing (full CRUD + reload)
drouting_list_rules,drouting_add_rule,drouting_update_rule,drouting_delete_ruledrouting_list_gateways,drouting_add_gateway,drouting_update_gateway,drouting_delete_gatewaydrouting_list_carriers,drouting_add_carrier,drouting_update_carrier,drouting_delete_carrier
Dialplan (number translation)
dialplan_list_rules,dialplan_add_rule,dialplan_update_rule,dialplan_delete_ruledialplan_reload,dialplan_translate
Carrier Routing & LCR
carrierroute_reload,carrierroute_dump,carrierroute_activate_host,carrierroute_deactivate_host,carrierroute_replace_hostlcr_dump,lcr_reload
B2BUA (back-to-back, SBC, SCA)
b2b_list_sessions,b2b_list_entities,b2b_terminate,b2b_bridge,b2b_trigger_scenario,b2b_sca_listtopology_hiding_list
Presence / PUA / BLF / MWI
presence_list_subscriptions,presence_refresh_watchers,presence_cleanuppua_list_publications,pua_publish
UAC / Upstream Registrations
uac_reg_list,uac_reg_reload,uac_reg_enable,uac_reg_disable,uac_reg_force_register
Mid-Registrar
mid_registrar_dump,mid_registrar_update
NAT Helper
nathelper_enable_ping,nathelper_stats
Event Interface (flatstore, rabbitmq, json, etc.)
event_list,event_list_subscribers,event_subscribe
Stateless (SL)
sl_stats
AVP Operations
avp_get,avp_set,avp_delete
Load Balancer
lb_list-- List load balancer destinationslb_reload-- Reload LB tablelb_set_status-- Enable/disable a destinationlb_resize-- Resize a resource
Call Center
cc_list_queue-- List call center queuecc_list_agents-- List agentscc_agent_login-- Login/logout an agentcc_reload_tool-- Reload call center configcc_status-- Call center statistics
Permissions
address_list-- List address permissionsaddress_add-- Add an address entryaddress_remove-- Remove an address entry
Security
pike_list_blocked-- List Pike-blocked IPspike_unblock_ip-- Unblock an IPsecurity_generate_ha1-- Generate HA1 digestssecurity_audit_config-- Audit a configuration
RTPEngine
rtpengine_show-- Show RTPEngine statusrtpengine_enable-- Enable/disable an RTPEngine instancertpengine_reload-- Reload RTPEngine config
Rate Limiting
ratelimit_status-- Show rate limit pipesratelimit_reset-- Reset a rate limit pipe
Cluster
cluster_status-- Cluster node statuscluster_topology-- Cluster topologycluster_send_command-- Send a cluster broadcast command
Tracing
trace_control-- Enable/disable SIP tracingtrace_start_live-- Start live SIP trace
Registrar
registrar_stats-- Registrar statistics
Accounting
acc_query_cdrs-- Query CDR recordsacc_query_missed-- Query missed call records
Diagnostics
get_process_list-- OpenSIPS process listget_memory_stats-- Shared/private memory statsget_uptime-- Server uptimeget_tcp_connections-- Active TCP connectionsopensips_version-- OpenSIPS version info
Configuration Engine — generation
cfg_generate-- Generate config from a scenario templatecfg_generate_iterative-- Generate and iteratively validatecfg_list_scenarios/cfg_get_scenario_params-- Scenario catalogcfg_start_session/cfg_generate_m4/cfg_preprocess/cfg_dry_run/cfg_save_session-- m4 / local.cfg override patterncfg_list_flags/cfg_compose_flags/cfg_build_from_flags-- Kamailio-compatibleWITH_*feature-flag composer (produces a flat config with only enabled branches materialised)
Configuration Engine — understanding
cfg_parse-- Parse config and extract structurecfg_validate-- Validate with theopensips -C -fbinarycfg_explain--kubectl explain-style inline reference for any module / modparam / function /$pv/ global / MI commandcfg_explain_route-- Analyse a specific route blockcfg_list_modules-- Module catalog (filter by version or category)
Configuration Engine — editing
cfg_diff-- Unified diff between two configurationscfg_diff_reference-- Diff against the reference scenario template to see what's "custom"cfg_edit-- Apply a natural-language edit and show the diffcfg_add_module-- Add a module (with optional modparams) into an existing cfgcfg_optimize-- Quick heuristic suggestions
Configuration Engine — linting & tuning
cfg_lint-- Run 20+ semantic lint rules (dependency gaps, deprecated params, missing anti-flood, etc.)cfg_list_lint_rules-- Rule catalogcfg_tune-- PGTune-style capacity tuner: describe workload → get override snippet with# why:comments
Configuration Engine — version migration
cfg_list_versions-- Supported versions and hop chain (2.4 → 3.0 → 3.1 → 3.2 → 3.3 → 3.4 → 3.5 → 3.6 fully populated; one anticipatory 3.6 → 4.0 hop)cfg_check_compat-- Check a config for modules unavailable in a target versioncfg_migrate_plan-- Terraform-plan-style dry-run: report every rewrite and every manual stepcfg_migrate-- Apply the migration, emit a unified diff, validate the result
MCP Resources
URI | Description |
| OpenSIPS version information |
| Server uptime |
| Running processes |
| Memory usage |
| All runtime statistics |
| Available modules |
| Module documentation |
| MI command documentation |
| Available config templates |
| Scenario template source |
| Deployment scenario catalog |
| Scenario detail and Docker config |
| Subscriber table snapshot |
| Dispatcher table snapshot |
MCP Prompts
Prompt | Description |
| Interactive OpenSIPS configuration builder |
| Performance tuning advisor |
| SIP architecture designer |
| Call failure troubleshooting guide |
| Registration failure troubleshooting |
| SIP/OpenSIPS error explainer |
| Security configuration auditor |
| SIP/VoIP concept educator |
| SIP ladder diagram analyzer |
| OpenSIPS route block explainer |
| SIP message parser and debugger |
| SIP capacity planning advisor |
| OpenSIPS C module development guide |
| Version migration assistant (natural-language plan) |
| Automated plan + apply workflow driving |
| Module comparison guide |
Architecture
+------------------+ +-------------------+ +------------------+
| MCP Client | | OpenSIPS MCP | | OpenSIPS |
| (Claude, etc.) |<--->| Server |<--->| (MI HTTP) |
+------------------+ | | +------------------+
| +-------------+ |
| | Tools (94) | | +------------------+
| +-------------+ | | Database |
| | Resources | |<--->| (MySQL/PG/ |
| +-------------+ | | SQLite) |
| | Prompts | | +------------------+
| +-------------+ |
| | Config Eng. | |
| +-------------+ |
| | Security | |
| +-------------+ |
+-------------------+
Layers:
MCP (FastMCP) -> Service (Tools/Resources/Prompts)
-> Infrastructure (MI Client, DB, Config Engine)
-> Version (3.4 / 3.6 strategies)
-> Security (RBAC, Validation, Audit)Supported OpenSIPS Versions
Version | Status | Notes |
2.4 | Migration source | Fully rule-backed source for cfg_migrate |
3.0–3.5 | Migration source | Each hop has fully-populated rule sets |
3.4 | Runtime supported | LTS — full MI and DB coverage |
3.6 | Runtime default + migration target | Latest stable LTS (3.6.4 as of 2026-04) |
4.0 | Anticipated migration target / experimental runtime | One advisory hop with master-branch MI renames; rules not yet validated against a stable build |
OpenSIPS jumped from 3.6 directly to 4.0 — there was no 3.7 release. The migrator chain reflects the actual release line (8 hops total: 7 fully populated to 3.6, plus 1 anticipatory hop to 4.0).
Development
# Setup
python -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
# Test
make test
# Lint and format
make lint
make format
make type-checkSee CONTRIBUTING.md for the full development guide.
Documentation
Start here:
Quickstart -- Zero to working proxy in 15 minutes (install → client setup → first prompts)
Tool Chooser -- "Which tool should I use for X?" — map goals to tools
FAQ / Troubleshooting -- Setup errors, runtime issues, common questions
Reference:
Architecture -- Layered design and component overview
Configuration Reference -- All environment variables and options
Security Model -- RBAC, validation, and audit logging
API Reference -- Complete tools, resources, and prompts listing
Deployment Guide -- Docker and bare metal deployment
Config Engine -- Scenario templates, linter, tuner, migration
Developer Guide -- Adding tools, templates, and testing
Claude Code skills and subagents (in .claude/skills/ and .claude/agents/):
Skills:
/opensips-quickstart,/opensips-config-build,/opensips-troubleshoot,/opensips-migrate,/opensips-security-audit,/opensips-tool-chooserSubagents:
opensips-architect,opensips-config-author,opensips-incident-responder,opensips-migrator,opensips-security-auditor,opensips-teacher
Examples
License
Apache License 2.0 -- Copyright 2026 Officering LLC
Available Tools
236 toolsacc_query_cdrsA
Query accounting CDRs with optional date range and caller/callee filters.
Parameters
date_from:
ISO-format start date/time filter (e.g. 2025-01-01T00:00:00).
date_to:
ISO-format end date/time filter.
caller:
Partial match on the caller field.
callee:
Partial match on the callee field.
Inputs: date_from, date_to, caller, callee, +2 more. Returns: object
acc_query_missedB
Query missed calls with optional date range and caller/callee filters.
Parameters
date_from:
ISO-format start date/time filter (e.g. 2025-01-01T00:00:00).
date_to:
ISO-format end date/time filter.
caller:
Partial match on the caller field.
callee:
Partial match on the callee field.
Inputs: date_from, date_to, caller, callee, +2 more. Returns: object
address_addB
Add a new address/permissions entry. Optionally triggers MI address_reload.
Inputs: ip (required), grp, mask, port, +4 more. Returns: object
address_listB
List address/permissions entries, optionally filtered by group.
Inputs: grp, limit, offset. Returns: object
address_removeB
Remove an address/permissions entry by ID. Optionally triggers MI address_reload.
Inputs: id (required), reload. Returns: object
avp_deleteC
Delete a global AVP.
Inputs: name (required). Returns: object
avp_getA
Get an AVP value (global scope only — per-transaction AVPs are not MI-accessible).
Parameters
name:
AVP name (e.g. $avp(my_var) — pass as my_var).
Inputs: name (required). Returns: object
avp_setC
Set an AVP value in the global scope.
Inputs: name (required), value (required). Returns: object
b2b_bridgeA
Bridge an existing B2B call to a new destination URI.
Useful for attended-transfer, failover, and music-on-hold scenarios.
Parameters
key:
Session key from b2b_list_sessions.
new_dst:
Target SIP URI to bridge to (e.g. sip:agent@domain).
Inputs: key (required), new_dst (required). Returns: object
b2b_list_entitiesA
List raw B2B entities (UAC/UAS/server) tracked by b2b_entities.
Inputs: no named parameters. Returns: object
b2b_list_sessionsA
List active B2B logic sessions (entity keys, state, legs).
Inputs: no named parameters. Returns: object
b2b_sca_listA
List active Shared Call Appearance (SCA) subscriptions.
Inputs: no named parameters. Returns: object
b2b_terminateA
Terminate a B2B session by its entity key.
Parameters
key:
The B2B session key as reported by b2b_list_sessions.
Inputs: key (required). Returns: object
b2b_trigger_scenarioA
Trigger a B2B scenario by id (defined in b2b_logic XML scenarios).
Parameters
scenario_id: Scenario id from the b2b_logic configuration. custom_hdrs: Optional custom SIP headers for the scenario invocation. params: Optional semicolon-delimited parameters (per-scenario).
Inputs: scenario_id (required), custom_hdrs, params. Returns: object
bench_quick_inviteB
Run a SIPp INVITE load test against TARGET.
Same shape as bench_quick_register but exercises the call path instead of the registration path.
Inputs: target (required), cps, duration_sec, out_dir, +1 more. Returns: object
bench_quick_registerA
Run a SIPp REGISTER load test against TARGET.
Captures pre/post snapshots of the proxy's state so the run is self-documenting. Default: 50 CPS for 30 s = 1500 registers, safe for any non-toy proxy.
Parameters
target:
host:port of the proxy under test.
cps:
Calls per second. Capped at 1000.
duration_sec:
Test duration. Capped at 600 (10 min).
out_dir:
Directory for the bench artefact JSON. Must be in safe
allowlist.
label:
Optional human tag (baseline, after-tune-shm, ...).
Inputs: target (required), cps, duration_sec, out_dir, +1 more. Returns: object
bench_summaryA
Aggregate every bench artefact in OUT_DIR into one summary.
Useful at the end of a tuning run: bench_summary produces a
table of (label, kind, CPS, success_rate, p50/p95 latency, shmem
delta) so you can see which knobs actually moved the needle.
Inputs: out_dir. Returns: object
carrierroute_activate_hostA
Activate a carrier route host (return to service).
Use this after maintenance to re-enable a previously deactivated gateway.
Inputs: carrier (required), domain (required), prefix (required), host (required). Returns: object
carrierroute_deactivate_hostC
Deactivate a carrier route host (drain, take out of service).
Inputs: carrier (required), domain (required), prefix (required), host (required). Returns: object
carrierroute_dumpA
Dump loaded carrier route tables (carriers, domains, prefixes, hosts).
Inputs: no named parameters. Returns: object
carrierroute_reloadA
Reload carrier route tables from the database.
Inputs: no named parameters. Returns: object
carrierroute_replace_hostA
Replace a host in carrier route (e.g. IP change).
Does not require a reload; applies at runtime.
Inputs: carrier (required), domain (required), prefix (required), host (required), +1 more. Returns: object
cc_agent_loginB
Log a call center agent in or out.
Parameters
agent_id:
The agent identifier (e.g. agent1).
state:
Login state: 1 to log in, 0 to log out.
Inputs: agent_id (required), state (required). Returns: object
cc_list_agentsA
List all call center agents and their current state.
Inputs: no named parameters. Returns: object
cc_list_queueA
List all calls currently queued in the call center.
Inputs: no named parameters. Returns: object
cc_reload_toolA
Reload call center agents and flows from the database.
Inputs: no named parameters. Returns: object
cc_statusA
Get a combined call center status: queue, agents, and flows.
Returns all three datasets in a single response for a complete overview of the call center state.
Inputs: no named parameters. Returns: object
cdr_auditA
Scan the acc table for gaps, duplicates, and duration anomalies.
Parameters
date_from, date_to: ISO timestamps bounding the audit window. Window is capped at 31 days; longer ranges should go through a warehouse / BI tool. max_gap_sec: Gap between consecutive CDRs that counts as a suspicious silence. Default 300s (5 min). long_call_threshold_sec: Duration above which a call is flagged as a potential stuck dialog. Default 14400s (4h).
Inputs: date_from (required), date_to (required), max_gap_sec, long_call_threshold_sec. Returns: object
cdr_count_by_codeB
Count CDRs by SIP response code within a window.
Quick-glance "what's the 5xx rate right now?" report — SQL-side aggregation so it is cheap even on large acc tables.
Inputs: date_from (required), date_to (required). Returns: object
cdr_summaryB
Volume and success-code breakdown over a window.
Parameters
date_from, date_to:
ISO timestamps bounding the report window.
group_by:
hour or day. Grouping is computed in Python so it works
across sqlite / mysql / postgres without backend-specific SQL.
Inputs: date_from (required), date_to (required), group_by. Returns: object
cfg_add_moduleA
Add a module (with optional modparams) to an existing OpenSIPS configuration.
Inserts the loadmodule directive and any modparam lines after the
last existing loadmodule or modparam block.
Parameters
config_content:
The existing OpenSIPS configuration text.
module_name:
The module to add (e.g. dispatcher, rtpengine).
params:
Optional dict of {param_name: value} to set via modparam.
Inputs: config_content (required), module_name (required), params. Returns: object
cfg_backupA
Copy config_path to a timestamped backup in a safe directory.
The backup name is <basename>.bak-YYYYMMDD-HHMMSS next to the
original (or in backup_dir when provided). Returns the backup
path so the caller can pass it to cfg_rollback later.
Parameters
config_path:
Absolute path of the existing config file. Must live in an
allowlisted directory (/tmp, /var/tmp, or $HOME;
override with OPENSIPS_MCP_OUT_DIR_ROOT).
backup_dir:
Optional destination directory for the backup copy. Default:
alongside the original file.
Inputs: config_path (required), backup_dir. Returns: object
cfg_build_from_flagsA
Build a flat opensips.cfg from a set of WITH_* flags.
Materialises the Kamailio-style flag vocabulary as a flat, no-#!ifdef
OpenSIPS config. Only the enabled blocks are inlined, so the output
reads cleanly without the nested-ifdef problem Kamailio users hit.
Parameters
flags:
Any combination of WITH_* flags (see cfg_list_flags).
site_params:
Optional site values (LISTEN, DB_URL, TLS_CERT_PATH,
TLS_KEY_PATH, RTP_PROXY_ADDR, CHILDREN).
validate:
Run the result through opensips -C when available.
Inputs: flags (required), site_params, validate. Returns: object
cfg_check_compatA
Check a config for incompatibilities with a target OpenSIPS version.
Reports modules that don't exist in the target version, deprecated
parameters, and silent-gotcha defaults. Pairs well with
:func:cfg_migrate_plan — run compat first to learn about removals,
plan next to see the rewrites.
Parameters
config_content: The OpenSIPS configuration text to check. target_version: The version you intend to run this config on.
Inputs: config_content (required), target_version (required). Returns: object
cfg_compose_flagsA
Resolve a set of WITH_* flags into modules and globals.
Does NOT produce a config — it answers the question "if I enable
these flags, what modules get loaded and which globals get set?".
Pairs with :func:cfg_build_from_flags for the full render.
Parameters
flags:
List of WITH_* flag names (case-insensitive).
Inputs: flags (required). Returns: object
cfg_diffA
Compute a unified diff between two OpenSIPS configurations.
Parameters
config_a: The first (original) configuration text. config_b: The second (modified) configuration text.
Inputs: config_a (required), config_b (required). Returns: object
cfg_diff_referenceA
Compare a config against the reference output for a scenario.
Shows which parts of the config are "custom" (differ from the canned scenario template). Useful for reviewing what your team has diverged on and would need to re-apply after upgrading the reference template.
Parameters
config_content:
The OpenSIPS configuration text to compare.
scenario:
The reference scenario name (see cfg_list_scenarios).
params:
Parameters used to render the reference.
Inputs: config_content (required), scenario (required), params. Returns: object
cfg_dry_runA
Safely iterate on a config: preprocess + validate, write nothing.
Intended as the LLM's tight feedback loop while building or editing a config — call this between successive renders to verify correctness before asking the user to save.
Parameters
main_m4:
Contents of opensips.cfg.m4.
local_m4:
Contents of local.cfg.m4.
Inputs: main_m4 (required), local_m4 (required). Returns: object
cfg_editA
Validate an edited OpenSIPS config and return a unified diff.
Does not write anything to disk — the caller chooses to persist after reviewing the diff.
Parameters
existing_cfg: The current on-disk configuration text. change_description: Short natural-language description of the intended change (e.g. "Increase children to 16 and enable TLS"). new_cfg: The LLM's rewritten configuration text.
Inputs: existing_cfg (required), change_description (required), new_cfg (required). Returns: object
cfg_explainA
Explain an OpenSIPS directive, module, function, or pseudo-variable.
kubectl explain equivalent. Accepts flexible topic syntax:
dialog— module overviewdialog.db_mode— specific modparamrecord_route— script function$ru— pseudo-variablelisten— global parameterlb_status— MI command (with version-rename history)
Parameters
topic: The directive / name to explain.
Inputs: topic (required). Returns: object
cfg_explain_routeA
Explain an OpenSIPS route block by identifying its components.
Analyzes the route code and identifies SIP methods handled, module functions called, conditions, and overall flow.
Parameters
route_code:
The route block code to explain (e.g. the body of a route{...}).
Inputs: route_code (required). Returns: object
cfg_generateA
Generate an OpenSIPS configuration from a predefined scenario template.
Renders the scenario template, then — by default — runs opensips -C -f
when the binary is on PATH and runs the static linter unconditionally.
The Summit deck claim "Every output runs through opensips -C -f" lives
here: this is the headline tool, and it owns the validate-on-render
contract. If the binary is unavailable, the result includes
validation: {"skipped": true, "reason": "opensips binary not found"}
so the caller can decide whether to proceed.
Parameters
scenario:
Name of the scenario (e.g. load_balancer, class4_sbc,
registrar_class5, webrtc_gateway, residential_pbx,
call_center, b2bua, sbc_with_rtpengine).
params:
Template parameters such as db_url, listen_ip, domain, etc.
validate:
When True (default), run opensips -C -f against the rendered
cfg. Skipped (with a reason) if the binary is not on PATH.
lint:
When True (default), run cfg_lint against the rendered cfg
and surface findings + suggestions. Pure Python, always available.
Inputs: scenario (required), params, validate, lint. Returns: object
cfg_generate_iterativeA
Generate an OpenSIPS config and iteratively validate it.
Generates a configuration from the given scenario, validates it using
opensips -C -f, and reports the results. If validation fails,
the raw output is included for diagnosis.
Parameters
scenario: The scenario template name. params: Template parameters. max_attempts: Maximum validation attempts (reserved for future auto-fix logic).
Inputs: scenario (required), params, max_attempts. Returns: object
cfg_generate_m4A
Render the m4 template pair (main + local) for a scenario.
The returned strings are intended to be persisted to opensips.cfg.m4
and local.cfg.m4 respectively, then preprocessed with m4.
Parameters
scenario:
Scenario name (load_balancer, class4_sbc, registrar_class5,
webrtc_gateway, base).
template_params:
Structural/Jinja2 template knobs (e.g. include_usrloc for base).
site_params:
Dict of m4 defines that will appear in local.cfg.m4 (e.g.
{"LISTEN": "udp:10.0.0.5:5060", "DB_URL": "mysql://..."}).
Inputs: scenario (required), template_params, site_params. Returns: object
cfg_get_scenario_paramsB
Get required and optional parameters for a configuration scenario.
Parameters
scenario:
The scenario name (e.g. load_balancer).
Inputs: scenario (required). Returns: object
cfg_lintA
Run the OpenSIPS semantic linter against a configuration.
Checks module dependencies, orphaned modparams, deprecated parameters,
missing mf_process_maxfwd_header, anti-flood absence, NAT/rtpengine
dependency gaps, and other common production pitfalls.
Parameters
config_content:
The OpenSIPS configuration text to lint.
include:
Optional allow-list of rule ids (e.g. ["OPS001", "OPS004"]).
exclude:
Optional deny-list of rule ids.
Inputs: config_content (required), include, exclude. Returns: object
cfg_list_backupsA
List available .bak-* backups for config_path.
Parameters
config_path:
Absolute path whose backups we want to enumerate. The basename
is used as the prefix filter.
backup_dir:
Optional directory to search. Defaults to the directory of
config_path.
Inputs: config_path (required), backup_dir. Returns: object
cfg_list_flagsA
List all available WITH_* feature flags.
The flag vocabulary is Kamailio-compatible so engineers migrating between the two projects keep the same mental model.
Inputs: no named parameters. Returns: object
cfg_list_lint_rulesA
Return all available lint rule ids.
Inputs: no named parameters. Returns: object
cfg_list_modulesA
List modules from the OpenSIPS module catalog.
Used by the wizard and by LLMs that need to check whether a module is available in a given version.
Parameters
version:
Filter to modules available in this OpenSIPS version (e.g. 3.6).
category:
Filter by category (core, transport, tls, db,
routing, auth, media, security, nat, ...).
Inputs: version, category. Returns: object
cfg_list_scenariosA
List all available OpenSIPS configuration scenarios with descriptions.
Inputs: no named parameters. Returns: object
cfg_list_versionsA
List supported OpenSIPS versions and the chain of migration hops.
Returns the known versions (from the migration rule tables) and the
summary of each hop (2.4 → 3.0 → ... → 4.0). Use this before
calling :func:cfg_migrate to confirm your source and target are
supported.
Inputs: no named parameters. Returns: object
cfg_m4_checkA
Diagnostic: check whether the m4 binary is available on the host.
Returns path, version, and availability status.
Inputs: no named parameters. Returns: object
cfg_migrateA
Migrate a configuration between OpenSIPS versions.
Walks the hop chain (2.4 → 3.0 → ... → 4.0) applying every
behaviour-preserving rewrite rule. Anything that cannot be safely
auto-applied is reported as a warning — the engineer must apply those
manually.
When validate is true and an opensips binary is available, the
migrated config is also run through opensips -C -f.
Parameters
config_content:
The current OpenSIPS configuration text.
from_version:
Current OpenSIPS version.
to_version:
Target OpenSIPS version.
validate:
Run the migrated config through opensips -C when available.
Inputs: config_content (required), from_version (required), to_version (required), validate. Returns: object
cfg_migrate_planA
Report what a version migration would do, without rewriting the config.
Terraform-plan equivalent. Returns the ordered list of hops, every
automatic rewrite, and every manual step / silent-gotcha warning.
Parameters
config_content:
The current OpenSIPS configuration text.
from_version:
Current OpenSIPS version (e.g. 3.4).
to_version:
Target OpenSIPS version (e.g. 3.6).
Inputs: config_content (required), from_version (required), to_version (required). Returns: object
cfg_optimizeB
Analyze an OpenSIPS configuration and suggest optimizations.
Checks for common issues: missing modules, redundant settings, performance tuning opportunities, and security recommendations.
Parameters
config_content: The OpenSIPS configuration text to analyze.
Inputs: config_content (required). Returns: object
cfg_parseA
Parse an OpenSIPS configuration and extract its structure.
Returns modules, modparams, listen addresses, routes, and global params.
Parameters
config_content: The full OpenSIPS configuration text to parse.
Inputs: config_content (required). Returns: object
cfg_preprocessA
Run m4 local.cfg.m4 opensips.cfg.m4 and optionally validate the output.
Parameters
main_m4:
Contents of opensips.cfg.m4.
local_m4:
Contents of local.cfg.m4 (site-specific define(...) lines).
validate:
If True and the opensips binary is present, run opensips -C -f
on the preprocessed output.
Inputs: main_m4 (required), local_m4 (required), validate. Returns: object
cfg_rollbackA
Restore backup_path to target_path and re-validate.
Before overwriting, the current contents of target_path (if any)
are themselves backed up so this tool is always reversible one more
step. If validate_after=True, the restored config is passed
through opensips -C; a validation failure does NOT undo the
restore (the caller asked for the restore explicitly), but the
result is surfaced so the caller can decide what to do next.
Parameters
backup_path:
Absolute path to the backup file to restore.
target_path:
Absolute path to overwrite. Must live in an allowlisted dir.
validate_after:
Run opensips -C on the restored file. Default True.
Inputs: backup_path (required), target_path (required), validate_after. Returns: object
cfg_save_sessionA
Persist the m4 pair to disk, optionally also writing the preprocessed cfg.
Writes {out_dir}/opensips.cfg.m4 and {out_dir}/local.cfg.m4. When
write_preprocessed is True, also runs m4 and writes
{out_dir}/opensips.cfg.
The output directory is validated — writes to system directories such as
/, /etc, /usr, /bin are rejected.
Parameters
out_dir:
Directory that will receive the files. Created if missing.
main_m4:
Contents for opensips.cfg.m4.
local_m4:
Contents for local.cfg.m4.
write_preprocessed:
If True, also run m4 and write opensips.cfg.
Inputs: out_dir (required), main_m4 (required), local_m4 (required), write_preprocessed. Returns: object
cfg_start_sessionB
Start an interactive m4 config-building session.
Takes a natural-language description (e.g. "I need a SIP load balancer for 4 backends") and returns the suggested scenario, required site parameters with descriptions, optional parameters, and a next-step instruction for the LLM.
Parameters
description: Natural-language description of the deployment the user wants.
Inputs: description (required). Returns: object
cfg_tuneA
Generate a PGTune-style capacity-tuned override snippet.
Ask the user five questions about their workload, return a small
paste-ready set of overrides with a # why: comment on every line.
Intended to be merged into an existing opensips.cfg — it is NOT a
full configuration.
Parameters
role:
One of edge, registrar, lb, b2bua, sbc,
presence, generic.
cpu_cores:
Physical or allocated CPU cores.
memory_mb:
RAM in megabytes available to OpenSIPS.
cps:
Expected peak calls per second.
concurrent_calls:
Expected peak concurrent calls.
tls:
True if TLS / WSS traffic is significant.
nat:
True if the deployment handles heavy NAT'd clients.
ha:
True if deployed active-active / clustered.
registrations:
Expected concurrent registered AoRs (registrar role).
Inputs: role, cpu_cores, memory_mb, cps, +5 more. Returns: object
cfg_validateA
Validate an OpenSIPS configuration by running opensips -C -f.
Parameters
config_content: The full OpenSIPS configuration text to validate.
Inputs: config_content (required). Returns: object
cleanup_test_dataA
Remove every subscriber matching <username_prefix>NNNN under
domain. Companion to gen_test_subscribers.
Parameters
domain:
Domain whose test users to remove.
username_prefix:
Same prefix used in generation.
confirm:
Must be True to actually delete. Default False returns
the dry-run count of matched rows.
Inputs: domain (required), username_prefix, confirm. Returns: object
cluster_drain_nodeA
Gracefully drain a cluster node ahead of maintenance.
Marks the node as disabled in the clusterer, then polls the local dialog count until it stops declining or the timeout fires. The caller is expected to stop the OpenSIPS process only after this tool confirms the dialog backlog is low.
This tool operates on the cluster metadata — it does NOT shut OpenSIPS down. Combine with your init-script / systemd unit to actually stop the service.
Parameters
cluster_id: Target cluster ID. node_id: Target node ID within the cluster. timeout_sec: Maximum time to wait for dialogs to drain. Clamped to 5..600. poll_interval_sec: Seconds between dialog-count samples. Clamped to 1..60.
Inputs: cluster_id (required), node_id (required), timeout_sec, poll_interval_sec. Returns: object
cluster_promote_nodeA
Bring a disabled cluster node back into active service.
Counterpart to cluster_drain_node: flips status back to OK so
peers resume sending traffic / replication to the node.
Inputs: cluster_id (required), node_id (required). Returns: object
cluster_send_commandA
Send an MI command to a specific node in the cluster.
Parameters
cluster_id:
The cluster ID.
node_id:
The target node ID within the cluster.
mi_cmd:
The MI command to execute on the remote node. Must be registered in
:data:MI_COMMANDS; the per-command permission is enforced so this
tool cannot be used to bypass mi.write controls.
mi_params:
Optional parameters for the MI command (JSON string).
Inputs: cluster_id (required), node_id (required), mi_cmd (required), mi_params. Returns: object
cluster_statusA
List all cluster nodes and their current status.
Inputs: no named parameters. Returns: object
cluster_sync_checkB
Report cluster sync health: capabilities, sharing tags, topology.
Combines clusterer_list, clusterer_list_cap and
clusterer_list_shtags into one view so an operator can answer
"is replication healthy across all nodes?" from a single call.
Inputs: cluster_id (required). Returns: object
cluster_topologyA
Show the cluster topology — how nodes are interconnected.
Inputs: no named parameters. Returns: object
db_table_backupA
Dump table to a timestamped JSON file under out_dir.
Parameters
table:
One of the supported OpenSIPS data tables (subscriber, dispatcher,
address, drouting_rule, drouting_gateway, drouting_carrier,
dialplan, load_balancer, domain, tls_mgm).
out_dir:
Directory to write into. Must be in the safe allowlist.
label:
Optional tag (e.g. before-bulk-import) added to the filename.
Inputs: table (required), out_dir (required), label. Returns: object
db_table_list_backupsA
List backup files for table under out_dir, newest first.
Inputs: table (required), out_dir (required). Returns: object
db_table_restoreA
Restore a table from a JSON backup produced by db_table_backup.
By default this is a dry-run. Pass confirm=True to actually
write. Pass truncate_first=True to delete every existing row in
the table BEFORE inserting the backup — needed when the backup is
intended to fully replace current state, but dangerous if you
misjudge the scope.
Parameters
backup_path:
Path to the JSON file produced by db_table_backup.
confirm:
Must be True to write. Default False returns the plan.
truncate_first:
If True, delete current rows before inserting from the
backup. Use when restoring a complete table snapshot.
Inputs: backup_path (required), confirm, truncate_first. Returns: object
diagnose_nat_flowA
Correlate dialog, contact, RTPEngine, and NAT state for one call.
The canonical "why is this call one-way audio?" triage tool. Pulls the four pieces of state that actually matter in NAT failures:
Dialog state from
dlg_list— is the call still active? which branches?User-location contact from
ul_show_contact(derived from the SIP From URI of the active dialog) — is the contact behind NAT and is the received field populated?RTPEngine state via
rtpengine_showandget_statisticsforrtpengine:— is media being relayed at all?Nathelper counters — is ping working? did the contact expire mid-call?
Returns a unified view plus a best-effort diagnosis string. A real engineer is still the best judge — this just consolidates the data so they don't have to bounce between four MI calls.
Parameters
callid: The SIP Call-ID of the problematic dialog.
Inputs: callid (required). Returns: object
dialplan_add_ruleB
Add a dialplan translation rule.
Parameters
dpid: Dialplan id (translation table id, referenced from script). match_op: Match operator — 0 = equal, 1 = regex (default), 2 = fnmatch. match_exp: Pattern to match against the input. subst_exp: Substitution regex (for match_op=1). repl_exp: Replacement string. pr: Priority — lower matches first.
Inputs: dpid (required), match_exp (required), subst_exp (required), repl_exp (required), +7 more. Returns: object
dialplan_delete_ruleC
Delete a dialplan rule by id.
Inputs: id (required), reload. Returns: object
dialplan_list_rulesB
List dialplan rules, optionally filtered by dialplan id.
Inputs: dpid, limit, offset. Returns: object
dialplan_reloadA
Reload dialplan rules from the database (triggers MI dp_reload).
Inputs: no named parameters. Returns: object
dialplan_translateA
Test a dialplan translation without dispatching a call.
Parameters
dpid:
Dialplan table id.
input:
Input string to translate (e.g. +14155551212).
Returns
The translated output string and matching rule metadata.
Inputs: dpid (required), input (required). Returns: object
dialplan_update_ruleC
Update a dialplan rule by id.
Inputs: id (required), dpid, match_op, match_exp, +8 more. Returns: object
dispatcher_addB
Add a new dispatcher destination. Optionally triggers MI ds_reload.
Inputs: destination (required), setid, flags, priority, +6 more. Returns: object
dispatcher_list_dbC
List dispatcher destinations from the database, optionally filtered by set ID.
Inputs: setid, limit, offset. Returns: object
dispatcher_removeB
Remove a dispatcher destination by ID. Optionally triggers MI ds_reload.
Inputs: id (required), reload. Returns: object
dispatcher_updateB
Update a dispatcher destination by ID. Optionally triggers MI ds_reload.
Inputs: id (required), destination, setid, flags, +7 more. Returns: object
dlg_endC
End (tear down) an active dialog.
Parameters
callid: The Call-ID of the dialog to terminate. from_tag: The From-tag of the dialog. to_tag: The To-tag of the dialog.
Inputs: callid (required), from_tag (required), to_tag (required). Returns: object
dlg_get_profilesA
Get dialog profile information.
Parameters
profile: Specific profile name to query size for. If omitted, lists all profiles.
Inputs: profile. Returns: object
dlg_listA
List active SIP dialogs.
Parameters
callid: Optional Call-ID to filter by a specific dialog. from_tag: Optional From-tag to further narrow the dialog lookup.
Inputs: callid, from_tag. Returns: object
dlg_push_varC
Push a variable into an active dialog.
Parameters
callid: The Call-ID of the target dialog. from_tag: The From-tag of the target dialog. var_name: Name of the dialog variable to set. var_value: Value to assign to the variable.
Inputs: callid (required), from_tag (required), var_name (required), var_value (required). Returns: object
docker_deploy_scenarioA
Return the docker-compose command to deploy a specific scenario.
Parameters
scenario:
Name of the scenario directory (e.g. load-balancer, class4-sbc).
Inputs: scenario (required). Returns: object
docker_list_scenariosA
List all available Docker deployment scenarios.
Returns the scenario names and their docker-compose file paths.
Inputs: no named parameters. Returns: object
docker_logsB
Retrieve recent logs from a Docker Compose service.
Parameters
service:
The service name to fetch logs for (default: opensips).
lines:
Number of recent log lines to return (default: 100).
Inputs: service, lines. Returns: object
docker_restartC
Restart a Docker Compose service.
Parameters
service:
The service name to restart (default: opensips).
Inputs: service. Returns: object
docker_statusA
Show the status of running Docker Compose services.
Executes docker compose ps and returns the output.
Inputs: no named parameters. Returns: object
domain_addC
Add a new SIP domain to the database.
Inputs: domain (required), attrs. Returns: object
domain_deleteC
Delete a SIP domain from the database.
Inputs: domain (required). Returns: object
domain_listA
List all locally-served SIP domains.
Inputs: no named parameters. Returns: object
domain_reloadA
Reload the domain table in OpenSIPS via MI command.
Inputs: no named parameters. Returns: object
drouting_add_carrierC
Add a new dynamic routing carrier. Optionally triggers MI dr_reload.
Inputs: carrierid (required), gwlist (required), flags, sort_alg, +4 more. Returns: object
drouting_add_gatewayC
Add a new dynamic routing gateway. Optionally triggers MI dr_reload.
Inputs: address (required), type, strip, pri_prefix, +6 more. Returns: object
drouting_add_ruleC
Add a new dynamic routing rule. Optionally triggers MI dr_reload.
Inputs: groupid (required), prefix (required), gwlist (required), priority, +8 more. Returns: object
drouting_delete_carrierC
Delete a dynamic routing carrier by id.
Inputs: id (required), reload. Returns: object
drouting_delete_gatewayC
Delete a dynamic routing gateway by gwid.
Inputs: gwid (required), reload. Returns: object
drouting_delete_ruleC
Delete a dynamic routing rule by ruleid.
Inputs: ruleid (required), reload. Returns: object
drouting_list_carriersC
List dynamic routing carriers.
Inputs: limit, offset. Returns: object
drouting_list_gatewaysC
List dynamic routing gateways.
Inputs: limit, offset. Returns: object
drouting_list_rulesA
List dynamic routing rules, optionally filtered by group or prefix.
Inputs: groupid, prefix, limit, offset. Returns: object
drouting_update_carrierC
Update a dynamic routing carrier by id.
Inputs: id (required), carrierid, gwlist, flags, +5 more. Returns: object
drouting_update_gatewayC
Update a dynamic routing gateway by gwid.
Inputs: gwid (required), address, type, strip, +7 more. Returns: object
drouting_update_ruleC
Update a dynamic routing rule by ruleid. Optionally triggers MI dr_reload.
Inputs: ruleid (required), groupid, prefix, gwlist, +9 more. Returns: object
ecosystem_healthA
Check health of all ecosystem components (OpenSIPS, Homer, RTPEngine, CP).
Returns a summary of which services are reachable and responsive.
Inputs: no named parameters. Returns: object
ecosystem_overviewA
Show an overview of configured ecosystem component URLs and available tools.
Returns the endpoint configuration for Homer, RTPEngine, and Control Panel along with a summary of ecosystem integration tools.
Inputs: no named parameters. Returns: object
event_listA
List all event types advertised by the running OpenSIPS.
Each entry includes event name (e.g. E_DLG_STATE_CHANGED) and
the modules currently registered as publishers.
Inputs: no named parameters. Returns: object
event_list_subscribersA
List active event subscribers, optionally filtered.
Parameters
event:
Event name to filter by (e.g. E_DLG_LOAD).
socket:
Transport socket URI to filter by (e.g. udp:1.2.3.4:8080).
Inputs: event, socket. Returns: object
event_subscribeB
Subscribe to an OpenSIPS event at a given transport socket.
Parameters
event:
Event name (e.g. E_DLG_STATE_CHANGED).
socket:
Destination URI for event delivery — format depends on the backend
(udp:host:port, rabbitmq:cluster/exchange, flatstore:/path).
expire:
Subscription TTL in seconds; 0 = permanent until OpenSIPS restart.
Inputs: event (required), socket (required), expire. Returns: object
fail2ban_ban_ipC
Manually ban an IP address in a fail2ban jail.
Inputs: ip (required), jail. Returns: object
fail2ban_jail_statusA
Get status of a specific fail2ban jail (e.g., opensips, opensips-auth).
Inputs: jail. Returns: object
fail2ban_list_bannedA
List all currently banned IPs for a fail2ban jail.
Inputs: jail. Returns: object
fail2ban_statusA
Get overall fail2ban status including list of active jails.
Inputs: no named parameters. Returns: object
fail2ban_unban_ipC
Unban an IP address from a fail2ban jail.
Inputs: ip (required), jail. Returns: object
fraud_add_ruleC
Add a new fraud detection rule.
Inputs: profileid (required), prefix, start_hour, end_hour, +9 more. Returns: object
fraud_delete_ruleA
Delete a fraud detection rule by rule ID.
Inputs: ruleid (required). Returns: object
fraud_list_rulesB
List fraud detection rules, optionally filtered by profile ID.
Inputs: profileid, limit, offset. Returns: object
fraud_reloadA
Reload fraud detection rules in OpenSIPS via MI command.
Inputs: no named parameters. Returns: object
fraud_update_ruleC
Update an existing fraud detection rule.
Inputs: ruleid (required), profileid, prefix, start_hour, +10 more. Returns: object
gen_test_dispatcherA
Provision a dispatcher set with a handful of test destinations.
Useful when you have a scenario template that uses
ds_select_dst(<set_id>, ...) but the dispatcher table is empty.
Parameters
set_id:
Dispatcher set id. Templates default to set 1.
destinations:
List of SIP URIs (e.g. ["sip:10.0.0.10:5060", "sip:10.0.0.11:5060"]).
dry_run:
Default True. Returns the plan without writing.
socket:
Optional outbound socket per row.
Inputs: set_id (required), destinations (required), dry_run, socket. Returns: object
gen_test_subscribersA
Generate N test subscribers with predictable usernames + passwords.
Useful for smoke-testing REGISTER paths and synthetic load against
a fresh deployment. Test users are named <prefix>NNNN with
deterministic passwords so a SIPp scenario can authenticate
without external state.
Parameters
domain:
Domain to provision under. Validated against validate_domain.
count:
Number to create. Cap: 500.
username_prefix:
Username prefix; default test produces test0001..testNNNN.
password_prefix:
Password prefix; defaults to secret so test passwords look
like secret0001.
dry_run:
Default True. Returns the plan without writing.
Inputs: domain (required), count, username_prefix, password_prefix, +1 more. Returns: object
get_memory_statsA
Show OpenSIPS memory usage details (shared and private memory).
Inputs: no named parameters. Returns: object
get_process_listA
List all running OpenSIPS processes with their PID and description.
Inputs: no named parameters. Returns: object
get_statisticsA
Retrieve OpenSIPS runtime statistics.
Parameters
group:
Statistics group (e.g. core, shmem, net, tm, sl,
usrloc, or all for everything).
name:
Optional specific statistic name within the group.
Inputs: group, name. Returns: object
get_tcp_connectionsB
List all active TCP and TLS connections.
Inputs: no named parameters. Returns: object
get_uptimeA
Show how long the OpenSIPS instance has been running.
Inputs: no named parameters. Returns: object
health_checkB
Run a comprehensive health check against the OpenSIPS instance.
Checks MI connectivity and retrieves basic runtime statistics.
Inputs: no named parameters. Returns: object
homer_correlate_callidB
Correlate a Call-ID across OpenSIPS hops (e.g. b2b, topology_hiding).
OpenSIPS rewrites Call-ID when topology_hiding's C flag is set or
when B2BUA links two legs. Homer's call-data endpoint returns
correlated sibling Call-IDs in the response when the search matches
a parent ID. This tool wraps that query.
Inputs: callid (required), from_ts (required), to_ts (required). Returns: object
homer_export_pcapA
Request a pcap export of the captured SIP messages for callid.
Homer 7 exposes /api/v3/export/call/messages/pcap. This tool
returns the pcap bytes (base64-encoded if the response is binary) so
an operator can save them and open in Wireshark. Useful when the
SIP ladder by itself is not enough and you need to see the raw
bytes (e.g. malformed headers, weird encoding).
Parameters
callid: Exact SIP Call-ID. from_ts, to_ts: Time window (Unix milliseconds).
Inputs: callid (required), from_ts (required), to_ts (required). Returns: object
homer_get_call_flowB
Retrieve the SIP call-flow (ladder diagram) for a call from Homer.
Parameters
call_id: The SIP Call-ID to fetch the flow for.
Inputs: call_id (required). Returns: object
homer_get_call_transactionA
Fetch the full SIP transaction ladder for a single Call-ID.
Returns every captured SIP message (requests and replies, across
proxies if multiple OpenSIPS nodes capture the same dialog) in
timestamp order. Hits the same /api/v3/search/call/data endpoint as
homer_search_calls_v7 — Homer doesn't have a dedicated "get
transaction" path; transaction reconstruction is done by setting
param.transaction.call=true and filtering by Call-ID.
Parameters
callid: Exact SIP Call-ID. from_ts, to_ts: Time window (Unix milliseconds). Homer is strict — the window must bracket the call start.
Inputs: callid (required), from_ts (required), to_ts (required). Returns: object
homer_healthA
Check whether the configured Homer instance is reachable.
Light probe — hits Homer's /api/v3/status endpoint. Useful when
deciding whether to even attempt a homer_search_calls_v7 call.
Returns {ok: false} (not raises) when Homer is not configured —
the absence of Homer must not break other tools.
Inputs: no named parameters. Returns: object
homer_search_callsC
Search SIP calls captured by Homer/SIPCAPTURE.
Parameters
caller: Filter by calling party (From user). callee: Filter by called party (To user). call_id: Filter by SIP Call-ID. time_from: Start of time range (ISO-8601 string). time_to: End of time range (ISO-8601 string).
Inputs: caller, callee, call_id, time_from, +1 more. Returns: object
homer_search_calls_v7A
Homer 7 native search with ms-precision timestamps + source-IP filter.
Counterpart to homer_search_calls (which takes ISO strings). Prefer
this when you are chasing a tight time window or want to filter by
source IP — useful when a carrier denies sending an INVITE.
Parameters
from_ts:
Start of the search window as a Unix timestamp in milliseconds.
Homer 7 expects millisecond precision on this endpoint.
to_ts:
End of the search window, Unix ms.
from_user:
Optional substring match on the SIP From user.
to_user:
Optional substring match on the SIP To user.
callid:
Exact SIP Call-ID to locate a specific dialog.
source_ip:
Optional source IP filter. Validated against ipaddress.ip_address.
limit:
Maximum rows to return (capped at 500).
Inputs: from_ts (required), to_ts (required), from_user, to_user, +3 more. Returns: object
homer_statsA
Retrieve Homer capture statistics (method distribution, totals).
Inputs: no named parameters. Returns: object
lb_listA
List all load balancer destination groups and their current state.
Inputs: no named parameters. Returns: object
lb_reloadA
Reload load balancer destinations from the database.
Inputs: no named parameters. Returns: object
lb_resizeC
Resize the capacity of a load balancer resource.
Parameters
group: The destination group ID. destination: The destination URI. resource: The resource name to resize. capacity: The new capacity value.
Inputs: group (required), destination (required), resource (required), capacity (required). Returns: object
lb_set_statusC
Enable or disable a load balancer destination.
Parameters
group: The destination group ID. destination: The destination URI. enabled: True to enable, False to disable.
Inputs: group (required), destination (required), enabled (required). Returns: object
lcr_dumpA
Dump LCR rules and gateways currently loaded in memory.
Inputs: no named parameters. Returns: object
lcr_reloadB
Reload LCR rules and gateways from the database.
Inputs: no named parameters. Returns: object
mid_registrar_dumpA
Dump mid-registrar AOR state (UAC→upstream mappings).
Inputs: no named parameters. Returns: object
mid_registrar_updateC
Force a mid-registrar record update for an AOR.
Inputs: aor (required). Returns: object
mi_executeA
Execute a known OpenSIPS MI command.
The command must be registered in :data:MI_COMMANDS. The specific
permission recorded on the catalogue entry (mi.read or mi.write)
is enforced in addition to the outer mi.execute scope, so a future
role with read-only MI access cannot use this tool to run a write
command.
Parameters
command:
The MI command name (e.g. ul_dump, lb_list).
params:
Optional parameters for the MI command.
Inputs: command (required), params. Returns: object
mi_list_commandsA
List all MI commands available on the connected OpenSIPS instance.
Inputs: no named parameters. Returns: object
module_add_mi_commandB
Add an MI command to an existing OpenSIPS module C source.
Inserts the MI command handler function and updates the mi_cmds array.
Parameters
module_source:
The existing C source code of the module.
command_name:
Name of the MI command to add (e.g. my_status).
command_description:
Description of what the MI command does.
Inputs: module_source (required), command_name (required), command_description. Returns: object
module_add_paramB
Add a module parameter to an existing OpenSIPS module C source.
Inserts the parameter variable declaration and updates the params array.
Parameters
module_source:
The existing C source code of the module.
param_name:
Name of the parameter (e.g. db_url).
param_type:
Parameter type: str or int.
default_value:
Default value for the parameter.
Inputs: module_source (required), param_name (required), param_type, default_value. Returns: object
module_compileB
Compile an OpenSIPS module from source.
Runs make modules for the specified module within the OpenSIPS
source tree.
Parameters
module_name:
Name of the module directory under modules/.
source_path:
Path to the OpenSIPS source tree root.
Inputs: module_name (required), source_path. Returns: object
module_generate_docsB
Generate README documentation for an OpenSIPS module.
Produces a markdown README with parameter, MI command, and usage documentation.
Parameters
name:
Module name.
description:
Module description.
mi_commands_json:
JSON array of MI commands (same format as module_scaffold).
params_json:
JSON array of parameters (same format as module_scaffold).
Inputs: name (required), description, mi_commands_json, params_json. Returns: object
module_scaffoldA
Generate a complete OpenSIPS C module skeleton.
Creates all necessary files for a new OpenSIPS module: C source, header, Makefile, README, and optional MI command implementations.
Parameters
name:
Module name (e.g. my_module). Must be a valid C identifier.
description:
Short description of the module's purpose.
mi_commands_json:
JSON array of MI commands. Each element should have name,
optional description, and optional params (list of strings).
Example: [{"name": "my_cmd", "description": "Does something", "params": ["arg1"]}]
params_json:
JSON array of module parameters. Each element should have name,
type (str or int), and optional default.
Example: [{"name": "db_url", "type": "str", "default": "mysql://localhost/opensips"}]
Inputs: name (required), description, mi_commands_json, params_json. Returns: object
module_validateA
Validate that an OpenSIPS module can be loaded.
Creates a minimal configuration and runs opensips -C to verify
the module's shared object can be loaded without errors.
Parameters
module_name:
Name of the module (without .so extension).
module_path:
Path to the directory containing the compiled .so file.
Inputs: module_name (required), module_path (required). Returns: object
nathelper_enable_pingA
Enable or disable NAT pinging for registered contacts.
NAT pings are OPTIONS or keepalive packets sent to registered UACs to keep their NAT bindings open. Disabling may cause contacts behind NAT to become unreachable after NAT timeout (~30s typical).
Parameters
enable:
True to enable pinging, False to disable.
Inputs: enable. Returns: object
nathelper_statsA
Return nathelper runtime statistics (ping counts, failure rate).
Useful for diagnosing NAT keepalive issues and sizing ping load.
Inputs: no named parameters. Returns: object
observability_categories_for_modulesA
Report which dashboard categories would fire for a given module set.
Useful for a "what would I get?" preview before calling the bundle
generator. Tolerates "dialog" or "dialog.so" indifferently.
Parameters
modules:
Module names with or without .so suffix.
Inputs: modules (required). Returns: object
observability_generate_bundleA
Build a full deployment bundle from a scenario.
The bundle pairs the cfg generator output with everything you need to actually observe the deployment::
opensips.cfg # rendered + prometheus.so injected
monitoring/
prometheus.yml # scrape config
docker-compose.monitoring.yml # Prometheus + Grafana stack
grafana/
provisioning/
datasources/datasource.yml
dashboards/dashboard.yml
dashboards/<deployment>-*.json # one per triggered category
README.mdCategories are picked from the cfg's loadmodule set, so the
dashboards never drift from the modules actually loaded.
Parameters
scenario:
A scenario name (see cfg_list_scenarios).
params:
Scenario parameters (same as cfg_generate).
deployment_name:
Slug used in titles, filenames, and Prometheus labels. Defaults
to the scenario name.
inject_prometheus:
Add prometheus.so + httpd.so to the cfg if absent.
Default True. Set False if you want to wire the exporter yourself.
prom_port:
Port the OpenSIPS httpd module listens on. Default 8888.
grafana_port:
External port to publish Grafana on. Default 3000.
opensips_host:
Hostname Prometheus scrapes. Default "opensips" to match
the docker network alias used by the project's ecosystem stack.
Inputs: scenario (required), params, deployment_name, inject_prometheus, +3 more. Returns: object
observability_generate_bundle_from_cfgA
Build a full deployment bundle from an arbitrary cfg (not a scenario).
Same output shape as observability_generate_bundle, but takes
the cfg directly. Useful when you've hand-edited a generator output
or migrated a legacy config and want the matching observability stack.
Inputs: config_content (required), deployment_name, inject_prometheus, prom_port, +2 more. Returns: object
observability_generate_dashboardsA
Render a scenario and synthesize matching Grafana dashboards.
The cfg is rendered (validating required params), the loaded module
set is parsed out, and one dashboard JSON is produced per triggered
panel category. Returns {dashboards: {filename: dashboard_dict}, ...}
plus the matched module list for transparency.
Parameters
scenario:
A scenario name (see cfg_list_scenarios).
params:
Same parameters you would pass to cfg_generate.
deployment_name:
Slug used in titles and filenames. Defaults to the scenario name.
extra_tags:
Additional Grafana tags appended to every dashboard.
Inputs: scenario (required), params, deployment_name, extra_tags. Returns: object
observability_generate_dashboards_from_cfgA
Synthesize Grafana dashboards from any OpenSIPS configuration text.
Parses loadmodule directives and emits one dashboard JSON per
triggered panel category. Works on hand-edited configs, the output of
cfg_generate, the output of cfg_build_from_flags — anything
that's syntactically a valid OpenSIPS cfg.
Parameters
config_content: Full OpenSIPS configuration text. deployment_name: Slug used in dashboard titles and filenames. extra_tags: Additional Grafana tags appended to every dashboard.
Inputs: config_content (required), deployment_name, extra_tags. Returns: object
observability_generate_grafana_provisioningA
Generate Grafana provisioning YAMLs (datasource + dashboard provider).
Drop the two files into Grafana's /etc/grafana/provisioning/
layout and Grafana auto-registers the Prometheus datasource and
auto-loads every JSON in dashboards_path on startup.
Parameters
prometheus_url: URL Grafana should use to query Prometheus. folder: Grafana folder name dashboards land in. dashboards_path: Filesystem path inside the Grafana container where dashboard JSONs are mounted.
Inputs: prometheus_url, folder, dashboards_path. Returns: object
observability_generate_monitoring_composeA
Generate a docker-compose snippet for a Prometheus + Grafana side-car.
The stack assumes the OpenSIPS service is on the same external Docker
network so Prometheus can scrape opensips:8888/prometheus.
Parameters
deployment_name:
Used in container names so multiple deployments coexist.
grafana_port, prometheus_port:
External ports to publish.
grafana_admin_password:
Set the Grafana admin password (default admin).
network:
External Docker network name. Create with
docker network create <name> before starting the stack.
Inputs: deployment_name, grafana_port, prometheus_port, grafana_admin_password, +1 more. Returns: object
observability_generate_prometheus_scrapeB
Generate a Prometheus scrape configuration.
Parameters
deployment_name:
Used as the instance label so dashboards can pivot on it.
opensips_host, prom_port:
Default target when targets is omitted.
metrics_path:
Path the prometheus.so module serves on (default /prometheus).
scrape_interval:
Prometheus scrape interval. Default 15s.
targets:
Optional list of [label, "host:port"] pairs for multi-instance
scraping. When omitted, a single target is generated from
opensips_host + prom_port.
Inputs: deployment_name, opensips_host, prom_port, metrics_path, +2 more. Returns: object
observability_inject_prometheusA
Add prometheus.so (and httpd.so if missing) to an OpenSIPS cfg.
Most scenarios do not load prometheus.so by default. This tool
adds the minimum lines needed for the Prometheus scrape config and
Grafana dashboards to actually work:
loadmodule "httpd.so"(if missing)loadmodule "prometheus.so"modparam("prometheus", "statistics", "")— expose all groupsmodparam("httpd", "port", <prom_port>)(if missing)
Idempotent: re-running on a cfg that already loads prometheus.so is
a no-op. Returns warnings if httpd is configured for a different
port (the scrape job needs to target that port instead).
Parameters
config_content: Existing OpenSIPS configuration text. prom_port: TCP port the httpd module should listen on. Default 8888.
Inputs: config_content (required), prom_port. Returns: object
observability_list_panel_categoriesA
List the Grafana panel categories the dashboard generator can emit.
Each category produces one dashboard JSON (when its trigger modules
are present in the target cfg). overview is always emitted.
Returns {categories: [{name, title, description, trigger_modules, panel_count}, ...]}.
Inputs: no named parameters. Returns: object
observability_save_bundleA
Persist a bundle's {path: content} map to disk.
The files map is the files field from
observability_generate_bundle (relative paths → text contents).
The destination out_dir is validated against the same allowlist
used by cfg_save_session — writes to system directories such
as /, /etc, /usr are rejected.
Parameters
out_dir: Absolute path to the bundle root. Created if missing. files: Mapping of relative path → file contents. overwrite: If False (default), refuse to write any file that already exists.
Inputs: out_dir (required), files (required), overwrite. Returns: object
opensips_cp_statusA
Check OpenSIPS Control Panel availability and status.
Inputs: no named parameters. Returns: object
opensips_versionA
Show the version string of the connected OpenSIPS instance.
Inputs: no named parameters. Returns: object
perf_hotspotsA
Quick "where is the pain?" summary.
Combines mem, get_statistics for core rates and error counters,
and a snapshot of process state. Returns a flat dict of the single most
useful numbers — shmem usage %, dialog count, tm active transactions,
error rate — so an operator (or LLM) can decide whether to dig deeper.
Inputs: no named parameters. Returns: object
perf_memory_reportA
Report OpenSIPS shared-memory and per-process package-memory usage.
shm is the pool OpenSIPS uses for cross-process state (dialogs,
usrloc, tm transactions). Running out triggers out of shm memory
errors that kill calls silently — this is the number to watch in
production. pkg is per-process private memory; a runaway module
leak shows up there first.
Inputs: no named parameters. Returns: object
perf_process_breakdownA
Group OpenSIPS child processes by role and count them.
Returns per-type counts (UDP receivers, TCP workers, timer processes,
HTTPD, MI workers, module-owned children) so an operator can spot
misconfiguration — e.g. children=4 on an 8-core machine, or a
missing tcp_workers that leaves TLS single-threaded.
Inputs: no named parameters. Returns: object
perf_sample_windowA
Sample runtime statistics at a fixed cadence over a short window.
Useful for "is the proxy under pressure right now?" snapshots. Pulls
the statistic groups at sample_count evenly-spaced points and
returns the raw time-series — the caller (or LLM) computes deltas.
For long-running observation use a real TSDB (Prometheus, Graphite)
via proto_hep or prometheus module — this tool is capped at
10 minutes / 60 samples to keep MCP responses manageable.
Parameters
duration_sec:
Total sampling window in seconds. Clamped to 1..600.
sample_count:
Number of equally-spaced samples. Clamped to 2..60.
groups:
Statistics group prefixes to capture. Default: core health
indicators (core, shmem, tm, sl, usrloc,
dialog, net).
Inputs: duration_sec, sample_count, groups. Returns: object
pike_list_blockedA
List IP addresses currently blocked by the Pike anti-flood module.
Inputs: no named parameters. Returns: object
pike_unblock_ipB
Remove an IP address from the Pike block list.
Parameters
ip: The IP address to unblock.
Inputs: ip (required). Returns: object
presence_cleanupA
Force cleanup of expired presence records (watchers + presentities).
Inputs: no named parameters. Returns: object
presence_list_subscriptionsA
List active presence subscriptions (the presence hash table).
Useful for BLF diagnosis: if a watcher's key phone button doesn't update, this table shows whether OpenSIPS sees the subscription.
Inputs: no named parameters. Returns: object
presence_refresh_watchersB
Force a NOTIFY to all watchers of a presence URI.
Parameters
pres_uri:
Presentity URI (e.g. sip:alice@example.com).
event:
Presence event package: presence, dialog, message-summary,
call-info, etc.
refresh_type:
0 = publish-refresh, 1 = performance-refresh.
Inputs: pres_uri (required), event, refresh_type. Returns: object
pua_list_publicationsB
List active PUA publications (OpenSIPS-as-publishing-UA).
Inputs: no named parameters. Returns: object
pua_publishB
Publish a presence event via PUA.
Parameters
pres_uri:
Presentity URI being published.
event:
Event package (presence, dialog, message-summary, ...).
body:
Body payload (PIDF, dialog-info XML, etc.).
content_type:
MIME type of body.
expires:
TTL in seconds.
Inputs: pres_uri (required), event (required), body (required), content_type, +5 more. Returns: object
ratelimit_resetA
Reset a rate limiting pipe counter to zero.
Parameters
pipe: The name of the pipe to reset.
Inputs: pipe (required). Returns: object
ratelimit_statusA
List all rate limiting pipes and their current counters.
Inputs: no named parameters. Returns: object
registrar_statsA
Retrieve registrar-related statistics.
Returns registration counts, expiry stats, and related counters from the OpenSIPS registrar and usrloc modules.
Inputs: no named parameters. Returns: object
reload_data_tablesA
Run the full chain of OpenSIPS data-table reloads.
Fires the appropriate *_reload MI commands in dependency order.
Missing modules (commands the loaded OpenSIPS does not expose) are
reported as skipped rather than failed so the orchestrator
succeeds on partial deployments.
Parameters
targets:
Optional whitelist of module names (e.g. ["dispatcher", "drouting"])
to limit the reload. Default: every reload in the dependency chain.
stop_on_error:
If True, abort on the first failing reload. Default False —
a drouting failure should not prevent a tls cert refresh.
delay_ms_between:
Small gap between MI calls so OpenSIPS timer processes are not
saturated by a burst. Default 100ms.
Inputs: targets, stop_on_error, delay_ms_between. Returns: object
reload_dry_runA
Preview which modules would be reloaded without actually reloading.
Useful when a user is about to press the "reload everything" button and wants to know the blast radius first.
Inputs: no named parameters. Returns: object
reload_singleA
Reload one data table and optionally snapshot stats before/after.
Parameters
module:
One of: permissions, domain, dialplan, dispatcher,
load_balancer, drouting, carrierroute, lcr, acc,
fraud_detection, uac_registrant, call_center, tls_mgm.
verify_stats:
If True, fetch the module's stats after reload so the caller can
visually confirm the new data is live.
Inputs: module (required), verify_stats. Returns: object
rtpengine_enableA
Enable or disable an RTPEngine instance.
Parameters
url:
The RTPEngine instance URL (e.g. udp:127.0.0.1:2223).
enabled:
True to enable, False to disable.
Inputs: url (required), enabled. Returns: object
rtpengine_instance_healthA
Return reachability and load for every configured RTPEngine instance.
rtpengine_show returns the raw list; this tool folds in the
disabled/enabled state and the most recent rtpengine: stats so
an operator can tell at a glance which instance to shed load from.
Inputs: no named parameters. Returns: object
rtpengine_reloadA
Reload the RTPEngine configuration from the database.
Inputs: no named parameters. Returns: object
rtpengine_sessionsB
List active RTPEngine media sessions.
Inputs: no named parameters. Returns: object
rtpengine_showA
Show all RTPEngine instances and their current status.
Inputs: no named parameters. Returns: object
rtpengine_statsA
Retrieve RTPEngine media proxy statistics.
Inputs: no named parameters. Returns: object
rtpengine_terminate_callA
Tear down the RTPEngine media session for callid.
Forces RTPEngine to delete the media flow. Use when a signalling bug
leaves a media session stranded without a dialog — SIP-level
dlg_end_dlg alone won't free the RTP ports.
Parameters
callid: Exact SIP Call-ID of the media session to terminate.
Inputs: callid (required). Returns: object
security_audit_configA
Perform a basic security audit on an OpenSIPS configuration snippet.
Checks for common misconfigurations and security anti-patterns.
Parameters
config_content: The OpenSIPS configuration text to analyse.
Returns
dict
Contains findings (list of issues) and summary counts.
Inputs: config_content (required). Returns: object
security_generate_ha1A
Generate HA1 and HA1B digest authentication hashes.
These hashes are used in the subscriber table for SIP digest auth.
Parameters
username: The SIP username. domain: The SIP domain. password: The plaintext password to hash.
Returns
dict
Contains ha1 (username:realm:password) and
ha1b (username@domain:realm:password) hex digests.
Inputs: username (required), domain (required), password (required). Returns: object
sip_generate_sipp_scenarioA
Produce a SIPp XML scenario and the exact shell command to run it.
Does NOT execute SIPp. Generates a scenario the operator can run on a host they control — the MCP process is the wrong place to originate load traffic from.
Parameters
kind:
One of options, register, invite_uas_ringing.
target:
Destination host[:port] for SIPp's -rsa argument.
from_user, to_user:
SIP user-part values the scenario will parameterize.
calls_per_sec, max_concurrent, total_calls:
Standard SIPp knobs (-r, -l, -m).
Inputs: kind (required), target (required), from_user, to_user, +3 more. Returns: object
sip_options_pingA
Send SIP OPTIONS probes to target and report response codes / RTT.
Pure Python UDP implementation — no external sipsak / sipp needed.
Useful for "can the MCP host even reach this next-hop?" checks.
Parameters
target:
host[:port]. Port defaults to 5060.
count:
Number of probes. Clamped to 1..20.
timeout_sec:
Per-probe response timeout. Clamped to 0.2..10.
from_uri:
SIP From URI. Default is a clearly synthetic probe identity so
the receiving endpoint can filter it out of real traffic analysis.
allow_public:
Must be True to target a public IP. Prevents accidental probing
of production carriers from a misconfigured MCP.
Inputs: target (required), count, timeout_sec, from_uri, +1 more. Returns: object
sipp_list_scenariosC
List available SIPp scenario files.
Parameters
scenarios_dir: Directory to search for .xml scenario files.
Inputs: scenarios_dir. Returns: result
sipp_load_testA
Run a SIPp load test against a target SIP server.
Parameters
target: Target host:port. cps: Calls per second to generate. duration: Test duration in seconds. scenario: Built-in scenario name (uac, uas) or path to .xml file.
Inputs: target (required), cps, duration, scenario. Returns: object
sipp_run_scenarioA
Run a SIPp XML scenario file against a target SIP server.
Parameters
scenario_file:
Path to the .xml SIPp scenario file.
target:
Target host:port (e.g. 192.168.1.1:5060).
options:
JSON string of extra CLI flags, e.g. {"-m": "10", "-r": "5"}.
Inputs: scenario_file (required), target (required), options. Returns: object
sl_statsA
Return stateless reply counters (per-code, totals).
Useful for:
Auth load (401/407 counts ≈ registration / INVITE auth volume)
Scanner detection (4xx spikes without legitimate traffic)
OPTIONS keepalive health (200 counts should match incoming OPTIONS)
Inputs: no named parameters. Returns: object
snapshot_captureA
Capture a structured state snapshot of the running OpenSIPS.
Parameters
out_dir:
Directory to save the JSON snapshot into. Must live in an
allowlisted dir (/tmp, /var/tmp, $HOME, or
OPENSIPS_MCP_OUT_DIR_ROOT).
label:
Optional human-readable tag (before-deploy, baseline,
post-incident). Becomes part of the filename.
extra_stat_groups:
Additional statistics group prefixes to include beyond the
defaults. E.g. ["rtpengine:", "stir_shaken:"].
Inputs: out_dir (required), label, extra_stat_groups. Returns: object
snapshot_diffA
Diff two snapshots and surface the deltas.
Computes per-stat deltas (after - before), flags processes that appeared or disappeared, dispatcher / load_balancer state changes, and cluster-node status flips. Unchanged keys are omitted by default.
Parameters
before_path:
Path to the baseline snapshot JSON.
after_path:
Path to the comparison snapshot JSON.
show_unchanged:
If True, include keys whose values did not change. Default
False — diff output stays compact.
Inputs: before_path (required), after_path (required), show_unchanged. Returns: object
snapshot_listA
List snapshots in a directory, newest first.
Useful when you've been taking periodic snapshots and need to find "the one from before yesterday's deploy."
Inputs: out_dir (required). Returns: object
sngrep_captureA
Capture SIP traffic using tcpdump (sngrep-compatible pcap output).
Parameters
interface:
Network interface to capture on (default: any).
duration:
Capture duration in seconds (max 300).
filter_expr:
BPF filter expression (default: port 5060).
output_file:
Path to write the pcap file. Must resolve under /tmp or
/var/tmp (overridable via OPENSIPS_MCP_CAPTURE_ROOTS).
Inputs: interface, duration, filter_expr, output_file. Returns: object
subscriber_bulk_exportA
Dump subscribers as CSV text.
Parameters
domain:
Optional domain filter.
include_ha1:
When True, include the ha1 digest column. Defaults to
False — credential material should not leave the DB by
accident.
limit:
Row cap (default and max: 5000).
Inputs: domain, include_ha1, limit. Returns: object
subscriber_bulk_importA
Import subscribers from a CSV blob with conflict handling.
Expected CSV header: username,domain,password[,email_address][,rpid].
Extra columns are ignored. Blank/comment (#-prefixed) lines are
dropped. Caps at 5000 rows per call — split larger imports.
Parameters
csv_content:
Full CSV text including header row.
dry_run:
When True (default), parse and validate but do not write. Use
this first so you can review the plan before committing.
on_conflict:
skip (default) — leave existing (username, domain) alone and
report them in skipped.
overwrite — update password / attributes on existing rows.
error — abort on first collision; nothing is committed.
Inputs: csv_content (required), dry_run, on_conflict. Returns: object
subscriber_createB
Create a new SIP subscriber with auto-generated HA1 hashes.
Inputs: username (required), domain (required), password (required). Returns: object
subscriber_deleteB
Delete a subscriber by ID.
Inputs: id (required). Returns: object
subscriber_getA
Get a single subscriber by username and domain.
Inputs: username (required), domain (required). Returns: object
subscriber_listC
List subscribers, optionally filtered by domain.
Inputs: domain, limit, offset. Returns: object
subscriber_updateC
Update a subscriber's password or attributes.
Inputs: id (required), password, email_address. Returns: object
tls_addC
Add a new TLS domain configuration.
Inputs: domain (required), type, method, verify_cert, +10 more. Returns: object
tls_check_all_expiryA
Bulk certificate expiry scan across ALL configured TLS domains.
Queries all TLS domains from the database, parses each PEM certificate, and returns a summary sorted by urgency (most urgent first):
Domain name
Expiry date
Days until expiry
Status: "ok" (>30 days), "warning" (7-30 days), "critical" (<7 days), "expired"
Inputs: no named parameters. Returns: object
tls_check_expiryC
Check certificate expiry for a TLS domain by parsing its PEM certificate.
Inputs: domain (required). Returns: object
tls_connectionsA
List active TLS connections with details.
Retrieves all TCP/TLS connections via the list_tcp_conns MI command,
then filters to show only TLS connections. Each entry includes:
Source/destination IP:port
Connection state
Protocol type (TLS vs TCP)
Alias info if available
Inputs: no named parameters. Returns: object
tls_deleteB
Remove a TLS domain configuration.
Inputs: domain (required). Returns: object
tls_handshake_errorsA
Focused view of TLS error statistics with severity assessment.
Pulls get_statistics("tls:") and returns ONLY error-related counters with:
Error type and count
Percentage of total attempts for that error category
Severity assessment (ok, elevated, high, critical)
Inputs: no named parameters. Returns: object
tls_listA
List all TLS domain configurations from the database.
Inputs: no named parameters. Returns: object
tls_reloadA
Reload TLS domains in OpenSIPS via MI command.
Inputs: no named parameters. Returns: object
tls_statisticsA
Pull TLS-specific statistics from the tls_mgm module.
Returns counters organised into categories:
Handshake stats: connect_success, connect_error, accept_success, accept_error
I/O stats: read_error, write_error
Uses the get_statistics("tls:") MI command internally.
Inputs: no named parameters. Returns: object
tls_updateC
Update an existing TLS domain configuration (certificate, key, etc.).
Inputs: domain (required), certificate, private_key, ca_list, +9 more. Returns: object
topology_hiding_listA
List active dialogs with topology hiding engaged.
Shows the mapped keys used to rewrite Via/Contact/Record-Route between ingress and egress legs.
Inputs: no named parameters. Returns: object
trace_controlA
Control SIP tracing — start, stop, or check status.
Parameters
action:
One of start, stop, or status.
trace_uri:
SIP URI for the trace destination (required for start).
Inputs: action (required), trace_uri. Returns: object
trace_start_liveB
Start a live SIP trace with optional filters.
Parameters
filter_caller: Filter traces to a specific caller URI or prefix. filter_callee: Filter traces to a specific callee URI or prefix. filter_ip: Filter traces to a specific source/destination IP address.
Inputs: filter_caller, filter_callee, filter_ip. Returns: object
uac_reg_disableA
Disable a UAC registration by AOR (stops re-registering).
Inputs: aor (required). Returns: object
uac_reg_enableC
Enable a UAC registration by AOR.
Inputs: aor (required). Returns: object
uac_reg_force_registerA
Force an immediate re-registration for an AOR.
Useful after credentials change or when an upstream carrier has invalidated the existing registration.
Inputs: aor (required). Returns: object
uac_reg_listA
List UAC registrations, optionally filtered by AOR or contact.
Shows carrier/trunk registrations OpenSIPS is currently holding. Each row includes state, last registration time, and expiry.
Inputs: aor, contact. Returns: object
uac_reg_reloadA
Reload UAC registrant configuration from the database.
Inputs: no named parameters. Returns: object
ul_add_contactA
Add a contact binding to the user location table.
Parameters
table:
The usrloc table (e.g. location).
aor:
The Address of Record (e.g. alice@example.com).
contact:
The contact URI to register (e.g. sip:alice@192.168.1.10:5060).
expires:
Expiration time in seconds (default: 3600).
q:
Contact priority / q-value (default: 1.0).
Inputs: table (required), aor (required), contact (required), expires, +1 more. Returns: object
ul_dumpA
Dump the entire user location table.
Parameters
brief: If True, return a compact summary instead of full details.
Inputs: brief. Returns: object
ul_remove_contactB
Remove a specific contact binding from the user location table.
Parameters
table:
The usrloc table (e.g. location).
aor:
The Address of Record (e.g. alice@example.com).
contact:
The contact URI to remove.
Inputs: table (required), aor (required), contact (required). Returns: object
ul_show_contactsA
Show registered contacts for an Address of Record (AOR).
Parameters
aor:
The Address of Record (e.g. alice@example.com).
table:
The usrloc table to query (default: location).
Inputs: aor (required), table. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 236 tools
Several tool clusters have genuinely fuzzy boundaries: sipp_load_test and sipp_run_scenario both execute SIPp scenarios, get_memory_stats vs perf_memory_report vs perf_hotspots all report the same shm/pkg memory data, cc_status already returns queue+agents+flows making cc_list_agents/cc_list_queue redundant, and rtpengine_show overlaps with rtpengine_instance_health. With 236 tools, an agent will frequently be unsure which of several near-duplicates to call.
The dominant module_verb convention is well followed across most domains (dialplan_add_rule, cluster_drain_node, tls_check_expiry, subscriber_bulk_import). However, there are systematic deviations: a get_* family (get_process_list, get_memory_stats, get_tcp_connections, get_statistics), a gen_* family (gen_test_subscribers, gen_test_dispatcher), and prefix-less outliers like health_check, cleanup_test_data, and sl_stats break the otherwise predictable pattern.
236 tools is extreme bloat — over 15x the upper bound of a well-scoped server and 4x the 'too many' threshold. Even though OpenSIPS administration is genuinely broad, this surface would be far more usable split into focused servers (config, routing, media, monitoring, subscriber management).
Within each subdomain the coverage is unusually deep: dialplan, dispatcher, drouting, subscriber, TLS, fraud, and config lifecycle all have full CRUD or create/get/update/delete/reload chains, plus extras like config migration planning, snapshots, and load-testing harnesses. Minor gaps exist (address has no update, domain lacks update, no direct service start/stop control) but agents can work around them.
Maintenance
Related MCP Connectors
SIP/VoIP/telecom grounding for AI agents: vendor docs, RFCs, STIR/SHAKEN, traces, configs.
ORDan's MCP server: calls, SMS and phone numbers for AI agents, under their owner's spend controls
Phone, SMS and email for AI agents. One remote MCP endpoint, OAuth login, zero install.
IP intelligence, network diagnostics, and AI service availability tools for AI agents.
Related MCP Servers
- AlicenseBqualityFmaintenanceAI-native control plane for Kubernetes and GitOps. Provides 30+ tools for service deployment, database provisioning, and log management via natural language.11MIT

oitvoip-mcp-serverofficial
AlicenseBqualityDmaintenanceA Model Context Protocol (MCP) server that provides AI agents with seamless access to the NetSapiens VoIP platform for user management, call analytics, and system administration.235MIT
dSIPRouter MCP Serverofficial
AlicenseNot gradedqualityDmaintenanceEnables AI assistants to manage dSIPRouter operations such as endpoint groups, carrier groups, inbound mappings, and call data retrieval through natural language.Apache 2.0- AlicenseNot gradedqualityBmaintenanceAn MCP server that enables AI assistants to manage Cisco Unified Communications Manager (CUCM) via AXL SOAP API, supporting provisioning, configuration, and database queries through natural language.199 npm7MIT