Skip to main content
Glama
TeamCity
by desty2k

readonly-mcp-teamcity

CI PyPI Python License: MIT

A read-only Model Context Protocol (MCP) server that lets AI agents explore a JetBrains TeamCity server: projects, build configurations, builds, logs, test results, problems, the queue, agents, and VCS changes. 29 tools, all of them reads.

Quick start

With uv installed, there is nothing else to install:

claude mcp add teamcity \
  -e TEAMCITY_BASE_URL=https://teamcity.example.com \
  -e TEAMCITY_TOKEN=your-token \
  -- uvx readonly-mcp-teamcity

uvx downloads the package into a cached, isolated environment and runs it. Python 3.11 or newer.

Read-only by design. This server only ever issues GET requests (a test fails if the HTTP client gains any other method). It cannot trigger, cancel, pin, tag, or modify anything in TeamCity, and every tool is annotated readOnlyHint. Scope the token read-only on the TeamCity side as a second line of defense: see Token. Read why I build read-only MCP servers.

Related MCP server: TeamCity MCP Server

Token

Create a token under Profile → Access Tokens. Set Permissions scope to Limit per project, pick the root project (or the projects the agent should see), and choose the Read-only preset. A token scoped that way cannot change anything even if a client misbehaves.

The Read-only preset selects the 14 VIEW_* permissions TeamCity marks as read-only, including global ones for agent details and server settings. A token never gets more than its owner has: with a token from a user who has only the Project viewer role, agent tools still work, and get_server_health reports metrics as unavailable but still returns health items. Checked against TeamCity 2026.2.1, where a preset token could not pin a build (HTTP 403).

Tools

Projects

Tool

Description

search_projects

Search projects by name, ID, or parent; browse the hierarchy.

get_project_details

Full project: child projects, build configs, parameters.

Build configurations

Tool

Description

search_build_configs

Structured search by project ID, build type ID, or paused state.

find_build_config

Fuzzy search across config names and project paths; flags ambiguous matches and lists matching projects.

get_build_config_details

Steps, triggers, parameters, VCS roots, settings, agent requirements.

get_build_config_branches

Branches a configuration can build, with default/active status.

Builds

Tool

Description

search_builds

Search build history filtered by config, project, status, branch, date, etc.

get_build_details

One build: status, agent, revisions, parameters, dependencies, test summary, tags.

get_build_statistics

Numeric metrics: durations, test counts, coverage, artifact size.

get_build_chain

Snapshot-dependency chain upstream (depends-on) and downstream (dependents).

Logs

Tool

Description

get_build_log

Paginated log retrieval with tail and line-range support.

search_build_log

Full-text search within a build log, with surrounding context.

Tests & problems

Tool

Description

diagnose_build

Why a build failed, in one call: problems, failed tests (unmuted first) and the error lines of the log.

get_build_test_results

Test occurrences for a build, filterable by status.

get_test_details

One test occurrence: failure reason, history brackets, mute/investigation state.

get_build_problems

Structured failure reasons (compilation, exit code, dependency).

search_investigations

Active investigations by project, build config, or assignee.

list_muted_tests

Currently muted tests with scope, who muted them, and why.

Infrastructure

Tool

Description

get_build_queue

Builds waiting for an agent, with wait reasons.

search_agents

Build agents with connection/authorization/enabled state and current build.

list_agent_pools

Agent pools with agent and project counts.

get_agent_details

One agent's state with enabled/authorized comments, last activity, current build.

list_build_artifacts

Browse a build's artifact tree (one directory level).

get_build_artifact_content

Read a text artifact (size-capped, line-paginated).

search_changes

VCS changes (commits) by build, config, project, committer, or since-commit.

get_change_details

One change with its full modified-file list.

resolve_teamcity_url

Parse a TeamCity web URL into project / build-config / build / change IDs.

get_server_info

TeamCity version, build number, and uptime.

get_server_health

Server health items + runtime metrics (CPU, memory, queue).

Install

Run it without installing (recommended):

uvx readonly-mcp-teamcity

Or install it into an environment:

uv tool install readonly-mcp-teamcity   # or: pip install readonly-mcp-teamcity

Docker

The image serves the streamable HTTP transport on port 8080. That transport has no authentication, so publish it on the host's loopback only:

docker run --rm \
  -e TEAMCITY_BASE_URL=https://teamcity.example.com \
  -e TEAMCITY_TOKEN=your-token \
  -p 127.0.0.1:8080:8080 \
  ghcr.io/desty2k/readonly-mcp-teamcity:latest

Configuration

All configuration comes from TEAMCITY_* environment variables.

Variable

Required

Default

Description

TEAMCITY_BASE_URL

yes

—

TeamCity server base URL (e.g. https://teamcity.example.com).

TEAMCITY_TOKEN

yes

—

REST API bearer token. See Token.

TEAMCITY_TIMEOUT

no

30

HTTP request timeout (seconds).

TEAMCITY_MAX_RETRIES

no

3

Retries for transient network / 5xx errors.

TEAMCITY_POOL_MAX_CONNECTIONS

no

20

httpx connection-pool size.

TEAMCITY_DEFAULT_LIMIT

no

50

Default page size for list tools.

TEAMCITY_MAX_LIMIT

no

500

Maximum page size.

TEAMCITY_INDEX_REFRESH_INTERVAL

no

300

Seconds between fuzzy-index refreshes. A failed load is retried sooner.

TEAMCITY_LOG_CACHE_MAX_BYTES

no

67108864

Size cap of the build-log cache. Logs of running builds are never cached.

TEAMCITY_LOG_CACHE_TTL

no

300

Seconds a finished build's log stays cached.

TEAMCITY_TRANSPORT

no

stdio

stdio, http, or sse.

TEAMCITY_HTTP_HOST

no

127.0.0.1

Bind address for http/sse. These transports have no authentication; a non-loopback address logs a warning.

TEAMCITY_HTTP_PORT

no

8080

Port for the http/sse transports.

TEAMCITY_LOG_FORMAT

no

json

json (production) or text (local dev).

TEAMCITY_LOG_LEVEL

no

INFO

DEBUG, INFO, WARNING, or ERROR.

See .env.example for a copy-paste template.

OpenTelemetry tracing

Tracing is opt-in and a no-op when disabled. The base install carries no OTel SDK; install the extra and enable it to export spans for each upstream TeamCity call:

pip install 'readonly-mcp-teamcity[otel]'
export TEAMCITY_OTEL_ENABLED=true
export TEAMCITY_OTEL_ENDPOINT=http://localhost:4317   # optional; falls back to OTEL_EXPORTER_OTLP_ENDPOINT

Variable

Default

Description

TEAMCITY_OTEL_ENABLED

false

Master switch for tracing.

TEAMCITY_OTEL_SERVICE_NAME

readonly-mcp-teamcity

service.name on every span.

TEAMCITY_OTEL_ENDPOINT

unset

OTLP collector endpoint.

TEAMCITY_OTEL_PROTOCOL

grpc

grpc or http/protobuf.

TEAMCITY_OTEL_CONSOLE_EXPORT

false

Also print spans to stderr.

MCP client configuration

Claude Desktop

Add to claude_desktop_config.json (~/Library/Application Support/Claude/ on macOS, %APPDATA%\Claude\ on Windows):

{
  "mcpServers": {
    "teamcity": {
      "command": "uvx",
      "args": ["readonly-mcp-teamcity"],
      "env": {
        "TEAMCITY_BASE_URL": "https://teamcity.example.com",
        "TEAMCITY_TOKEN": "your-token"
      }
    }
  }
}

Claude Code

See Quick start.

How it compares

As of November 2026. Check the projects themselves before relying on this.

This server

JetBrains MCP (2026.1+)

Daghis 2.12.12

itcaat 1.0.8

kate-horshchar 1.4.0

Writes

none, by construction

personal builds in the default safe mode; more with brave mode

trigger and cancel by default; a full mode the agent can switch on

trigger, cancel, pin, tag; no off switch

none, GET only

Runs where

your machine

the TeamCity server (/app/mcp)

your machine

your machine

your machine

Find a config from a vague name

fuzzy index over names and project paths

not documented

locator only

locator partial match

no

Paste a TeamCity URL

modern, short and legacy URLs

not documented

no

no

no

Sources: JetBrains AI agent integration, JetBrains/teamcity-mcp, Daghis/teamcity-mcp, itcaat/teamcity-mcp, kate-horshchar/teamcity-qa-mcp. The JetBrains plugin README and its help page disagree on whether brave mode is on by default; check your server's setting.

Use the built-in server if you want an agent to act on TeamCity. Use this one when the agent should only look.

Example questions

  • "Find the deploy pipeline for the auth service."

  • "Show me the last 10 builds for Proj_Build and which failed."

  • "Why did build 12345 fail?" (one diagnose_build call)

  • "What tests failed in that build, and what was the error?"

  • "Search the build log for OutOfMemoryError."

  • "Which agents are offline right now?"

  • "What's in the build queue and why is it waiting?"

  • "Who is investigating the failures in the Payments project?"

  • "What commits went into this build?"

  • "Resolve https://teamcity.example.com/buildConfiguration/Proj_Build/12345."

Development

git clone https://github.com/desty2k/readonly-mcp-teamcity
cd readonly-mcp-teamcity
uv sync --all-extras
uv run pytest --cov
uv run ruff check .
uv run ruff format --check .

A live smoke test against a real server (reads credentials from .env):

uv run python tests/smoke_test.py

License

MIT — see LICENSE.

Related MCP Connectors

Related MCP Servers

  • A
    license
    A
    quality
    A
    maintenance
    Enables AI coding assistants to interact with JetBrains TeamCity CI/CD server through natural language commands. Supports triggering builds, monitoring status, analyzing test failures, and managing build configurations directly from development environments.
    32
    3,258 npm
    31
    MIT
  • A
    license
    Not graded
    quality
    D
    maintenance
    Enables management of TeamCity CI/CD operations through natural language commands, including builds, projects, tests, artifacts, and more.
    21 npm
    MIT
  • A
    license
    Not graded
    quality
    C
    maintenance
    Provides read-only access to Jenkins, enabling users to inspect project and environment build status, list builds, retrieve console logs and test reports, diagnose failed builds, and search jobs via natural language, without triggering or changing builds.
    1,290 npm
    MIT
  • A
    license
    A
    quality
    C
    maintenance
    Enables AI coding agents to check Travis CI build status, identify failing jobs, and read job logs, with read-only access enforced via GET-only requests and a limited tool set.
    4
    MIT