Skip to main content
Glama
openITCOCKPIT MCP Server

openITCOCKPIT MCP Server

Official

openITCOCKPIT MCP Server

An MCP server that puts an openITCOCKPIT monitoring instance in front of an LLM client: what is broken and why, what someone already handles, what changed - and, off by default, the tools that act on it.

CAUTION

Use at your own risk. This server lets a language model read and - with write tools enabled - change your monitoring configuration. A model can misunderstand a request or pick the wrong call. Review what it proposes before you approve it, and start with read-only access. The software is provided "as is", without warranty or liability of any kind; see the MIT License.

  • Requires openITCOCKPIT 5.6 or newer (compatibility).

  • 42 tools, 20 read-only and 22 that change something.

  • Write tools are off by default and are not registered until you enable them.

  • Names, never IDs. Tools take hostnames, template names and container paths; the server resolves them.

  • Scope-checked writes. References are validated against the target container before anything is sent, which openITCOCKPIT's own API does not do.

Quickstart

cp .env.example .env          # fill in the two secrets, see Configuration
docker compose up --build

Point your client at http://localhost:8000/mcp with the bearer token from your .env. Compose reads that same file for the published port, so setting OITC_PORT there moves both sides at once.

Three settings matter to start: MCP_AUTH_TOKEN (what clients present to this server), OITC_BASEURL and OITC_APIKEY (what this server presents to openITCOCKPIT). Everything else has a default - all of it.

Related MCP server: Overwatch MCP

What you can ask it

Ask

What answers it

"What is broken right now?"

get_problem_overview - causes separated from their consequences

"Why is db-01 critical?"

get_service_health, investigate_problem

"Is anyone on it already?"

the same tools: downtime and acknowledgement come with the state

"What happened during the night shift?"

get_shift_summary

"Why did I get no alert for web01?"

explain_notification

"Take web01 out until Monday"

schedule_downtime (write)

"Which templates could web-05 use?"

get_allowed_elements_for_container

Every tool by name, with its parameters and what it costs in context: docs/tools.md, generated from the code. What a result looks like and what happens when you write: docs/using-the-tools.md.

Limiting an instance

Running a monitoring instance is not one job, so a surface small enough to be easy would be too small to be useful. Instead the server keeps every tool the work needs and a toolset hands one agent its share: around ten tools, complete for that job and nothing beyond it. A tool that was never registered cannot be called at all.

OITC_TOOLSETS=health          # find things and see how they are doing
OITC_TOOLSETS=operations      # acknowledge, maintenance windows, check now
OITC_TOOLSETS=lifecycle       # take things out of the monitoring, or delete them

The shipped sets are health, operations, lifecycle, patch, reporting, catalog, onboarding, config and provisioning. Sets are defined in a file you can replace: docs/toolsets.md.

Measuring a model

oitc-mcp-eval asks operator questions against your instance and checks whether a model answers them from the tools instead of inventing. It keeps every run in SQLite, so models and changes stay comparable.

oitc-mcp-eval --model your-model --samples 3     # against the configured instance
oitc-mcp-eval --compare                          # newest run of every model
./scripts/eval-throwaway.sh --model your-model   # instance created and removed for you

docs/evals.md.

Documentation

Configuration

every setting, and which secret goes where

Connecting a client

HTTP and stdio, with working examples

Installation

Docker, from source, MCP Registry

Tools

every tool, generated from the code

Working with the tools

result shapes, container scope, read-modify-write

Toolsets

limiting an instance, writing your own set

Skills and prompts

the material served alongside the tools

Measuring a model

the eval, its cases, and reading the results

Security

what this server can reach, and what it refuses

Architecture

how the pieces fit together

Tool design

the rules a new tool follows

openITCOCKPIT API notes

the behaviour this server works around

Versioning

version numbers and compatibility

Development

checks, tests, running from a checkout

Security in one paragraph

The server holds one openITCOCKPIT credential and never sends it to a client. Clients authenticate with their own token; in delegated mode each request carries a short-lived user token instead, and the server acts as that user with their permissions. Write tools stay unregistered until you enable them. The details, including what to give the openITCOCKPIT user: docs/security.md.

License

MIT.

Available Tools

20 tools
explain_notificationExplain NotificationA
Read-onlyIdempotent

Why a host or service did or did not notify: its settings, contacts and state checked the way Naemon checks them - disabled, downtime, soft state, host down, state not subscribed, acknowledged, no contact - and the notifications sent recently. Use it for "why did I get no alert for web01" or "who gets notified for Backup on db01".

Inputs: hostname (required), hours, servicename. Returns: object

find_downtimesFind DowntimesA
Read-onlyIdempotent

Find running and planned downtimes of hosts and services, by host, service or comment. Cancelled and expired ones are left out. Use it for "is web01 in maintenance" or "what is planned for tonight".

Inputs: host, kind, limit, comment, +2 more. Returns: object

find_hostsFind HostsA
Read-onlyIdempotent

Find hosts by name, state, container or host group, and count them per state. For unreachable hosts it names the down hosts they sit behind. Use it for "which hosts are down" or "how many hosts does tenant X have". For how one known host is doing, use get_host_health.

Inputs: name, limit, state, container, +3 more. Returns: object

find_noisy_checksNoisy ChecksA
Read-onlyIdempotent

Checks that cause noise and what to do about each: flapping services, the services that notified most, and problems that have lasted without anyone acknowledging them or scheduling a downtime. Use it for "what is noisy", "why do we get so many alerts" or "what should we clean up".

Inputs: hours, limit, container, hostgroup, +1 more. Returns: object

find_pending_updatesPending UpdatesA
Read-onlyIdempotent

Hosts with pending updates, with the package names and versions for each.

Covers every update by default; security_only narrows it to the security ones. Naming each package costs one API request, capped by max_packages_per_host; the update count itself is always exact.

Inputs: limit, security_only, max_packages_per_host. Returns: Rows plus enough context to know whether they are all of them.

find_servicesFind ServicesA
Read-onlyIdempotent

Find services by host, name, state, container, host group or flapping, and count them per state. Use it for "which services are critical", "what runs on host web01" or "which disk checks warn". For how one known service is doing, use get_service_health.

Inputs: host, name, limit, state, +5 more. Returns: object

forecast_metricForecast MetricA
Read-onlyIdempotent

When a measured value reaches its warning or critical threshold, from a straight line through its history. Reports the rate of change, the date and how well the line fits, per metric of the check. Use it for "when will the disk on db01 be full" or "what is filling up".

Inputs: hostname (required), servicename (required), hours, metric. Returns: object

get_allowed_elements_for_containerAllowed Elements for ContainerA
Read-onlyIdempotent

List the host templates, contacts, contact groups, timeperiods, host groups, etc. that are actually visible from a given container - i.e. the values a Create* tool for that object_type would accept there. openITCOCKPIT restricts every such reference to the target container's own scope (the container plus its descendants, plus a few legacy tenant-wide exceptions); values outside that scope are rejected. Call this BEFORE a create call whenever you are unsure a name is visible in the target container, instead of guessing and retrying on error. object_type must be one of: host, hosttemplate, servicetemplate, hostgroup, contactgroup, servicetemplategroup, contact. For hostgroup/contactgroup/servicetemplategroup/contact, container_name is the intended parent container (the object being created doesn't have its own container yet) - the result always includes 'legal_parent_containers' (the container types allowed to hold that object type), plus a members list (contacts/servicetemplates/timeperiods) only if container_name already resolves to a legal parent. container_name defaults to the root container if not given.

Inputs: object_type (required), container_name. Returns: object

get_availability_reportAvailability ReportA
Read-onlyIdempotent

How available a host or service was over a period, computed from its recorded state changes. Reports the percentage with and without agreed maintenance, and how long it spent in each state. Use it for "how available was shop01 last month" or "did we keep the SLA".

Inputs: hostname (required), days, servicename. Returns: object

get_configuration_statusConfiguration StatusA
Read-onlyIdempotent

Whether the monitoring engine runs the configuration as it stands: when it was last exported, what was changed since, what is configured but not monitored yet, and whether an export is running or even possible. Use it before and after a configuration change.

Inputs: no named parameters. Returns: object

get_container_treeContainer TreeA
Read-onlyIdempotent

Get the organizational structure (containers: tenants, locations, nodes) starting at the given container, including which hosts, host groups and service groups live directly under it. Leave container_name at 'root' for the top-level structure; without access to root that is the top-most containers you can see.

Inputs: container_name. Returns: object

get_host_healthHost HealthA
Read-onlyIdempotent

How one host is doing: state and since when, what likely explains a problem (a down parent, a downtime, an acknowledgement), which hosts depend on it, its services by state, and recent state changes. Use it for "how is web01", "why is web01 down" or "what does switch01 take down with it".

Inputs: hostname (required), hours. Returns: object

get_impactImpactA
Read-onlyIdempotent

What a host or service carries, before you disable, delete or take it out of the monitoring: its services, the hosts that depend on it, and the groups, maps and reports that name it. Use it to say what a change would affect.

Inputs: hostname (required), servicename. Returns: object

get_problem_overviewProblem OverviewA
Read-onlyIdempotent

What is broken and what is only a consequence: down hosts and the unreachable hosts and services behind them, the remaining unhandled service problems grouped by service, and what is already in a downtime or acknowledged. Use it first for "what is broken" or "give me an overview".

Inputs: groups, container, hostgroup. Returns: object

get_service_configService ConfigurationA
Read-onlyIdempotent

How a service is configured and where each value comes from: what it sets itself and what it follows its servicetemplate for. Field names are the ones update_service takes, so a value can be changed by the name it is reported under. It reports the configuration, never the current state or check results.

Inputs: hostname (required), servicename (required). Returns: object

get_service_healthService HealthA
Read-onlyIdempotent

How one service is doing: state and since when, what likely explains a problem (its host down, a downtime, an acknowledgement), and recent state changes. Use it for "why is Backup on db01 critical".

Inputs: hostname (required), servicename (required), hours. Returns: object

get_shift_summaryShift SummaryA
Read-onlyIdempotent

A handover of the last hours: problems that began, what someone acknowledged or put in a downtime, how many notifications went out, and what is still open from before. Use it for "what happened during the night shift" or "give me a handover".

Inputs: hours. Returns: object

investigate_problemInvestigate ProblemA
Read-onlyIdempotent

What happened around a problem: when it began, whether it happened before and how long it lasted then, which other hosts and services failed in the same minutes, and which configuration changes and exports came before. Works on the current problem, or on the last one when the object is fine again. Use it for "why did web01 go down" or "has Backup on db01 failed before".

Inputs: hostname (required), days, servicename, around_minutes. Returns: object

list_catalogConfiguration CatalogA
Read-onlyIdempotent

List templates, commands, contacts, groups or time periods by name, sorted by name, to find the exact name another tool needs. The total counts every match; items lists the first. Use it for "which service templates exist for disks" or "what is the host group of tenant X called".

Inputs: kind (required), name, limit. Returns: object

list_installed_softwareInstalled SoftwareA
Read-onlyIdempotent

Software installed on a host, from the openITCOCKPIT agent's inventory. OS is auto-detected (Linux, Windows, macOS).

A host carries hundreds to thousands of packages. Pass name_filter to search by package name, or only_updatable=True for just the outdated ones. For updates across the whole estate use find_pending_updates.

A host with no agent inventory at all raises, rather than returning zero rows.

Inputs: hostname (required), limit, name_filter, only_updatable. Returns: Rows plus enough context to know whether they are all of them.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.2/5.0

    Scored across 20 tools

    Disambiguation5/5

    Each tool targets a distinct aspect of monitoring—from finding hosts/services to explaining notifications, forecasting metrics, and investigating problems. There is minimal overlap; even similar tools like get_host_health and get_problem_overview serve different scopes (single object vs. global overview).

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern (get_, find_, list_, explain_, forecast_, investigate_). No mixed conventions or vague verbs; compound names like get_allowed_elements_for_container are still predictable.

    Tool Count4/5

    With 20 tools, the count is slightly above the typical 3-15 range, but it is justified for a comprehensive monitoring server covering search, health, configuration, reporting, and updates. The tools are well-scoped and not redundant, though a few could be merged (e.g., find_hosts and find_services are similar in style).

    Completeness4/5

    The tool surface covers the core monitoring lifecycle—discovering, diagnosing, and reporting—but lacks write operations (create/update/delete) for hosts or services, despite get_allowed_elements_for_container referencing create calls. This is a notable gap for full configuration management, but for read-only analysis it is sufficient.

    Maintenance

    ActivityStale
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      A
      maintenance
      A Model Context Protocol (MCP) server for Grafana. This provides access to your Grafana instance and the surrounding ecosystem.
      3,688
      Apache 2.0
    • A
      license
      Not graded
      quality
      D
      maintenance
      An MCP server that enables querying logs and metrics from Graylog, Prometheus, and InfluxDB 2.x. It provides tools for executing Lucene log searches, PromQL queries, and Flux queries directly within MCP-compatible clients.
      MIT
    • A
      license
      A
      quality
      F
      maintenance
      MCP server for monitoring health of MCP servers, GitHub Actions, GitLab pipelines, and HTTP endpoints with health history, TLS expiry, assertions, diagnostics, and alerts.
      22
      97 npm
      2
      MIT