woodpecker-mcp
Read-only MCP server for Woodpecker CI that allows AI agents to check pipeline status and retrieve step logs across multiple self-hosted instances.
README
woodpecker-mcp
MCP (Model Context Protocol) server for Woodpecker CI. Gives AI agents and MCP clients read-only visibility into pipeline status and logs across multiple self-hosted Woodpecker instances — so a code reviewer (human or agent) can check whether CI is green and pull the failing step's output without leaving the conversation.
Built against the Woodpecker v3 REST API (verified on v3.16.0). No write operations: the server never restarts, approves, or cancels pipelines.
Tools
| Tool | Purpose |
|---|---|
list_instances |
List configured instances; check: true verifies connectivity + token per instance |
list_repos |
Repositories the token can access on an instance |
list_pipelines |
Recent pipelines for a repo, filterable by branch / event / status |
get_pipeline |
One pipeline with workflows, steps, exit codes, and failed_step_ids |
get_step_logs |
Log output of a step (tailed, default last 100 lines) |
Repos are addressed as owner/name or by numeric Woodpecker repo id. get_pipeline accepts number: "latest" (optionally with branch).
Configuration
Instances are configured via environment variables — one URL/token pair per instance:
# Single instance (named "default"). WOODPECKER_SERVER (the official CLI var) also works.
WOODPECKER_URL=https://ci.example.com
WOODPECKER_TOKEN=<personal access token>
# Or one pair per instance — the name becomes the `instance` argument on every tool:
WOODPECKER_PROD_URL=https://ci.example.com
WOODPECKER_PROD_TOKEN=...
WOODPECKER_STAGING_URL=https://ci.staging.example
WOODPECKER_STAGING_TOKEN=...
Tokens come from <server>/user/cli-and-api on each Woodpecker instance.
The server also loads a .env.woodpecker file from its working directory if present (MCP clients launch stdio servers with the project as working directory). Real environment variables take precedence over file values.
Instances with self-signed certificates: point Node at your CA with NODE_EXTRA_CA_CERTS=/path/to/ca.pem.
Usage with Claude Code
Register in a project's .mcp.json (or claude mcp add):
{
"mcpServers": {
"woodpecker": {
"command": "npx",
"args": ["-y", "github:RikiGomes/woodpecker-mcp"]
}
}
}
Alternatively, clone it once and point at the local build:
{
"mcpServers": {
"woodpecker": {
"command": "node",
"args": ["/path/to/woodpecker-mcp/dist/index.js"]
}
}
}
(after npm install && npm run build in the clone).
Development
Node >= 22.18 runs the TypeScript sources directly (type stripping):
npm install
npm run dev # run the server from src/
npm run check # typecheck
npm test # vitest
npm run build # emit dist/
Everything is plain ESM; relative imports use explicit .ts extensions and the build rewrites them to .js.
Testing with the MCP Inspector
The MCP Inspector is the quickest way to poke at the tools by hand:
npm run build
npx @modelcontextprotocol/inspector node dist/index.js -e WOODPECKER_URL=https://ci.example.com -e WOODPECKER_TOKEN=<token>
The Inspector does not pass your shell environment to the server it launches. Prefixing the command (WOODPECKER_URL=… npx @modelcontextprotocol/inspector …) or exporting beforehand sets the variable on the Inspector, not on this server — which then starts with no instances configured. Use one of:
- the
-e KEY=VALUEflags shown above — note they go afternode dist/index.js, not before; - the Environment Variables fields in the Inspector's connection pane (then Connect / Restart);
- a
.env.woodpeckerfile in the directory you launch from (the name is exact — a plain.envis not read).
Scripted checks work too, via the Inspector's CLI mode:
npx @modelcontextprotocol/inspector --cli node dist/index.js --method tools/list
npx @modelcontextprotocol/inspector --cli node dist/index.js --method tools/call --tool-name list_instances
Design notes
- Read-only by contract. Intended for reviewer agents; there is deliberately no restart/approve/cancel surface.
- Log tailing. Woodpecker returns a step's entire stored log in one response;
get_step_logstails (default 100 lines, max 2000) to protect the caller's context window. Exit-code entries are labelled[exit code]. - Slug resolution.
owner/nameis resolved via/api/repos/lookup/…once and cached per process. - Sub-path hosting. Instance base URLs may include a sub-path (
WOODPECKER_ROOT_PATHinstalls).
License
MIT — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.