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., "@crowdsec-ops-mcp show top offending IPs in the last 24h"
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.
crowdsec-ops-mcp is a local MCP server for CrowdSec operations.
It exposes CrowdSec decisions, alerts, summaries, and safe single-IP action proposals to MCP clients. It is intentionally CrowdSec-only: it does not connect to VictoriaMetrics, VictoriaLogs, Grafana, Snort, reverse proxies, or Docker.
Supported runtime reads use CrowdSec LAPI. Decision reads use a bouncer API key. Alert reads require optional CrowdSec machine credentials because bouncer keys can only read decisions. Actual cscli reads or cscli execution are not supported by the MCP today; write tools only generate cscli command text for an operator to review and run outside the MCP if appropriate.
Related MCP server: crowdsec-mcp
Check whether the CrowdSec backend is reachable.
Inspect active decisions and recent alerts for one IP.
Summarize recent CrowdSec activity.
Find top offending source IPs.
Generate scenario-tuning proposals from repeated alert patterns.
Prepare audited single-IP ban, allow, or unban commands for manual review.
Write tools do not execute CrowdSec changes. They validate a single IP, prepare a plausible cscli command for an operator to review, append the prepared intent to the JSON Lines audit log, and return executed=false.
See ONBOARDING.md for installation, deployment, MCP client configuration, first tool calls, safety notes, and troubleshooting.
See docs/decision-inventory-example.md for example decision_inventory tool calls.
See docs/decision-gap-report-example.md for example decision_gap_report tool calls.
crowdsec_health(include_sample_counts=false)
inspect_ip(ip, window?)
security_summary(window?)
top_offenders(window?)
recent_crowdsec_decisions(window?)
decision_inventory(action?, origin?, scenario?, country?, asn?, ip?, limit?, expiring_soon_hours?, long_lived_days?)
recent_crowdsec_alerts(window?)
decision_gap_report(window?, repeat_threshold?, noisy_scenario_threshold?, expiring_soon_hours?, limit?)
suggest_scenario(window?)
unban_ip(ip, reason?, execute=false)
allow_ip(ip, duration?, reason, execute=false)
ban_ip(ip, duration?, reason, execute=false)
Variable | Purpose |
| CrowdSec LAPI base URL. Required for supported read operations. |
| CrowdSec LAPI key for decision reads. |
| Optional CrowdSec machine ID for read-only alert list access. |
| Optional CrowdSec machine password for read-only alert list access. |
| Command name/path used only when formatting prepared |
| Default lookback window, defaults to |
| JSON Lines audit trail for prepared write intents, defaults to |
| Python log level, defaults to |
ONBOARDING.md: user installation and first-use guide
CHANGELOG.md: version history
docs/decision-inventory-example.md: example decision_inventory tool calls
docs/decision-gap-report-example.md: example decision_gap_report tool calls
docs/snort-crowdsec-scenarios.md: example CrowdSec scenarios for Snort-derived alert patterns
docs/agent-usage.md: agent prompt patterns and investigation guidance
docs/development.md: local development workflow
docs/roadmap.md: project roadmap
docs/release-notes/v0.3.0.md: release notes for v0.3.0
docs/release-notes/v0.2.1.md: release notes for v0.2.1
docs/RELEASE.md: release process
docs/pull-request-rules.md: pull request rules
docs/release-notes-template.md: release notes template
Prepare and audit a potential cscli command to add a temporary allow decision for one IP. The MCP does not execute it.
Inputs: ip (required), reason (required), execute, duration.
Prepare and audit a potential cscli command to add a CrowdSec ban decision for one IP. The MCP does not execute it.
Inputs: ip (required), reason (required), execute, duration.
Report read-only CrowdSec MCP backend health, configuration presence, and exposed capabilities without exposing secrets.
Inputs: include_sample_counts.
Compare active CrowdSec decisions with recent alerts and return read-only attention findings.
Inputs: limit, window, repeat_threshold, expiring_soon_hours, +1 more.
Summarize active CrowdSec decisions with filters, grouped counts, expiry views, and representative rows.
Inputs: ip, asn, limit, action, +5 more.
Inspect CrowdSec decisions and CrowdSec alerts for one IP.
Inputs: ip (required), window.
Return recent CrowdSec alerts and alert visibility status. LAPI alert lists require machine auth.
Inputs: window.
Return active CrowdSec decisions.
Inputs: window.
Summarize recent CrowdSec decisions and alerts.
Inputs: window.
Suggest a CrowdSec scenario proposal from repeated CrowdSec patterns.
Inputs: window.
Return top source IPs by recent CrowdSec alert volume.
Inputs: window.
Prepare and audit a potential cscli command to delete a CrowdSec decision for one IP. The MCP does not execute it.
Inputs: ip (required), reason, execute.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 12 tools
Most tools have distinct purposes (e.g., inspect_ip vs. top_offenders vs. decision_inventory), but security_summary and recent_crowdsec_decisions/alerts overlap slightly in scope. Descriptions clarify the differences, so ambiguity is low.
Tool names use consistent snake_case and are descriptive, but the pattern varies between verb_noun (allow_ip, suggest_scenario), adjective_noun (top_offenders, recent_crowdsec_decisions), and noun_noun (crowdsec_health, decision_inventory). Still predictable and readable.
12 tools is well-scoped for a CrowdSec operations server. Each tool serves a clear purpose without bloat, covering health, inspection, listing, summarizing, gap analysis, and command preparation.
The tool set covers core read-only operations (alerts, decisions, health, inspection) and includes command preparation for allow/ban/unban. Missing execution capability is intentional, but the server could benefit from a tool to list scenarios or more detailed alert filtering. Minor gaps.
MCP server for mandates, delegation, policy-gated execution, credential grants, and audit.
A paid remote MCP for hosted MCP server, built to return verdicts, receipts, usage logs, and audit-r
MCP server for AI agents to plan, verify, and deploy Cloudflare-native apps.
An MCP server that let you interact with Cycloid.io Internal Development Portal and Platform
