Skip to content
MCP server

CICD AIops

Governed AI-ops for self-managed GitLab and Gitea: pipelines, runners, artifacts, RCA

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI
Runs
Locally
Version
0.14.0
Listing updated
today

Similar MCP servers

About CICD AIops

CICD AIops is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.

Built for teams who run their own GitLab or Gitea server, not the SaaS versions, it speaks GitLab REST v4 and Gitea API v1 directly with credentials encrypted at rest. The README is candid that it has been tested against mocked HTTP responses only, with no recorded end-to-end run on a live server, and points to a doctor command as the fastest live check.

Four analyses are the reason it exists: pipeline_failure_rca classifies failed pipelines (test failure, dependency or network, runner timeout, OOM, script error) from job trace tails; runner_health_rca finds offline, stale or paused runners and per-tag saturation; artifact_storage_bloat_analysis ranks projects by repo and artifact bytes with a reclaimable estimate; and stale_work_audit finds idle merge requests, dead branches and protection gaps. Every flag carries its numbers.

It does not decide whether a write is allowed; the token's permissions do. What it guarantees is a record: every MCP and CLI call writes an audit row, with a runaway guard against loops, undo entries for pause or resume runner and branch protection, and a before-state for retries, cancels and artifact deletes. Gitea has no runner or artifact-delete equivalents, and the tool returns an explanatory error there. Kubernetes deploy state is out of scope.

Install

Run in a terminal

claude mcp add cicd-aiops -- uvx cicd-aiops

Add to claude_desktop_config.json

{
  "mcpServers": {
    "cicd-aiops": {
      "command": "uvx",
      "args": [
        "cicd-aiops"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "cicd-aiops": {
      "command": "uvx",
      "args": [
        "cicd-aiops"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "cicd-aiops": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "cicd-aiops"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "cicd-aiops": {
      "command": "uvx",
      "args": [
        "cicd-aiops"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "cicd-aiops": {
      "command": "uvx",
      "args": [
        "cicd-aiops"
      ]
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Tools this server provides (10)

Taken from the project documentation; the running server may differ.

pipeline_failure_rca
Classifies recent failed pipelines from failed-job trace tails
runner_health_rca
Finds offline, stale or paused runners and tag saturation
artifact_storage_bloat_analysis
Ranks projects by repo and artifact bytes with reclaimable estimate
stale_work_audit
Finds idle merge requests, stale branches and protection gaps
retry_pipeline
Retries a pipeline (GitLab)
cancel_pipeline
Cancels a pipeline (GitLab)
pause_runner
Pauses a runner (GitLab)
resume_runner
Resumes a runner (GitLab)
delete_artifacts
Deletes job artifacts (GitLab)
update_branch_protection
Updates branch protection settings

What it does

Pipeline failure RCA
Classifies recent failed pipelines with the matched evidence and a suggested action.
Runner health analysis
Flags offline, stale or paused runners and tags where queued jobs outrun runners.
Artifact storage bloat
Ranks projects by repo and artifact size and estimates reclaimable bytes.
Stale work audit
Finds idle merge or pull requests, inactive branches and weak branch protection.
Audit and undo
Every call is logged locally; reversible writes store a replayable inverse.
Dry-run writes
CLI writes have `--dry-run` and double confirmation.

Good for

  • Find out why a self-hosted GitLab pipeline keeps failing
  • Spot wedged or offline CI runners
  • Work out where artifact storage went and what can be deleted
  • List stale merge requests and unprotected default branches
  • Retry or cancel pipelines with an audit trail

Pricing

Open source

MIT licensed

CICD AIops FAQ

Does it work with GitLab.com?

No. The README says GitLab.com and Gitea Cloud SaaS are out of scope; it targets self-managed instances.

Can it block dangerous writes?

No. It records every call but leaves permission to the token's scope on the server.

Has it been tested against a live server?

The README says only against mocked HTTP responses so far; `cicd-aiops doctor` checks a live connection.

What does Gitea lack?

No runner fleet, retry, cancel, pause or artifact-delete equivalents; those return an explanatory error.

How do I install it?

`uv tool install cicd-aiops` or `pip install cicd-aiops`, or as a Claude Code plugin.

Is CICD AIops free?

Yes. CICD AIops is open source, so you can use and self-host it for free.

How do I add CICD AIops to Claude Code?

Run this command in your terminal: claude mcp add cicd-aiops -- uvx cicd-aiops

What are the best alternatives to CICD AIops?

The closest alternatives to CICD AIops are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.

Where this comes from

  • Description: written from the project's README.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers