depot-mcp
Provides read-only tools for inspecting GitHub Actions jobs running on Depot runners — listing jobs by repo, status, conclusion and time, fetching an individual job, searching log lines, retrieving log context around a match, listing past runs of a job to spot flaky tests, and getting a job's test results summary.
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., "@depot-mcpwhy did the last CI run for my repo fail?"
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.
depot-mcp
A small, read-only MCP server for the Depot API.
Depot gives agents a CLI, but no MCP server. This server lets an agent read Depot CI failures, job logs, GitHub Actions jobs on Depot runners, and container builds. The agent does not need the depot CLI in its sandbox.
Transport: Streamable HTTP, stateless. One endpoint:
POST /mcp.Auth: Each request sends the caller's own Depot API token as
Authorization: Bearer <token>. The server sends that token to Depot and stores nothing.Scope: Read-only. No tool can start, cancel, retry or delete anything.
Tools
Depot CI
Tool | Depot RPC | Use |
|
| Find runs by repo, status, PR, SHA or trigger |
|
| Metadata and status of one run (no jobs or attempts) |
|
| Find workflows |
|
| A workflow with its jobs |
|
| Why a run, workflow, job or attempt failed |
|
| Markdown job summary |
|
| Job attempt logs, with tail, step, stderr and text filters |
|
| Artifact metadata for a run |
GitHub Actions on Depot runners
Tool | Depot RPC | Use |
|
| Find jobs by repo, status, conclusion and time (default: last 24 hours) |
|
| One job |
|
| Search log lines |
|
| Lines around a search match |
|
| Past runs of one job, to find flaky tests |
|
| Test results for a job |
Container builds
Tool | Depot RPC | Use |
|
| Projects in the organization |
|
| Builds for a project |
|
| One build |
|
| Build steps, with errors |
|
| Logs for one build step |
Typical flow to debug a failed Depot CI run:
ci_list_runswithstatus: ["failed"],repoandpr.ci_diagnose_failurewithtargetType: "run"and the run ID.ci_get_job_logswith an attempt ID from the diagnosis. AddstderrOnlyorcontainsto make the output smaller.
ci_get_run does not return jobs or attempt IDs. Use ci_diagnose_failure or ci_get_workflow to get them.
gha_list_jobs uses a 24-hour window when you do not set startAt or endAt. Depot's own default is 30 days, and that can time out on large organizations.
Related MCP server: depot-mcp
Known issues
gha_search_logsreturns500 internalfrom Depot. A direct call toapi.depot.devgives the same error, so the fault is in the Depot API. Because of this,gha_get_log_contextis not tested: it needs alineIdfromgha_search_logs.get_build_step_logscan return no lines, also for steps that ran for some seconds. The Depot API itself returns{}for these steps. CheckhasLogsanderroringet_build_stepsfirst.
Tokens
Use a Depot organization token (Organization settings → API tokens). Organization and user tokens can use every tool. Project tokens can use only the container build tools.
If you use a user token and belong to more than one organization, send X-Depot-Org: <org-id> as well.
Run
pnpm install
pnpm build
pnpm start # listens on PORT (default 3000)Or with Docker:
docker build -t depot-mcp .
docker run -p 3000:3000 depot-mcpEnvironment variables:
Name | Default | Description |
|
| HTTP port |
|
| Bind address |
|
| Depot API base URL |
|
| Timeout for each Depot API call |
GET /healthz returns {"status":"ok"}.
Put the server behind TLS when you deploy it. The token goes in a request header.
Connect a client
Claude Code:
claude mcp add --transport http depot https://<your-host>/mcp \
--header "Authorization: Bearer $DEPOT_TOKEN"Any other MCP client that supports Streamable HTTP and custom headers works the same way.
Development
pnpm dev # run with reload
pnpm test # vitest
pnpm typecheck
pnpm smoke PostHog/posthog # call the tools on a running server with DEPOT_TOKEN from .envThe server calls Depot's Connect RPC API with plain JSON over fetch. Request field names come from Depot's protobuf definitions in depot/proto and depot/cli.
This server cannot be deployed
Maintenance
Related MCP Connectors
Inspect Depot builds, CI runs, job logs and Actions runners; retry or cancel CI runs.
Read-only access to a Lumin project's logs, metrics, uptime checks, alerts and infrastructure.
Read-only access to Auralogs production logs: search logs, inspect errors, review AI analyses.
Read-only access to your CodeMouse accounts, repositories, and AI pull-request reviews.
Related MCP Servers
- FlicenseAqualityBmaintenanceEnables read-only inspection of pipeline jobs, pods, configurations, and logs from the dev EKS cluster and log store.8-
- AlicenseAqualityAmaintenanceProvides coding agents read-only access to Depot's CI failure diagnoses, container build forensics, run history, cache effectiveness, registry contents, and usage data through MCP tools.2851 npm1-
- AlicenseNot gradedqualityBmaintenanceEnables inspecting customer-data sources, destinations, connections, and warehouses through Segment's Public API with read-only access.482 npmMIT

Litport MCP serverofficial
AlicenseNot gradedqualityCmaintenanceEnables coding agents and chat clients to inspect proxy tokens, usage, pools, geo targeting, and build ready-to-use proxy connection URLs via the Litport account API, while remaining read-only.733 npmMIT