Callahan CI is a single Go binary that runs your pipelines on your laptop, reads diffs as they're built, and posts AI review and security findings back to your PRs. Four agents — Architect, Debugger, Reviewer, Analyst — do the parts you'd usually skip.
Callahan runs CI/CD pipelines locally and reviews the diff on every build. The Architect writes your YAML, the Debugger explains failures, the Reviewer posts structured findings on every PR, the Analyst parses Trivy and Semgrep into plain English.
One binary. Four agents. Pipelines as plain shell processes. Callahan is the small CI/CD tool you'd write yourself if you had a free weekend — built openly, MIT licensed, on your machine.
Callahan models a pipeline as a straight line of deterministic stages. Each runs natively on the host — no containers required — and emits a structured event stream the agents subscribe to.
Pipelines, AI review, security scanning, PR comments, and a dashboard — all in the binary. There's a marketplace tab in some other tools where you'd otherwise wire up half this list.
Paste your existing workflow. It runs. Prefer to describe it in plain English? The Pipeline Architect writes the YAML for you and commits it.
name: production on: [push, pull_request] jobs: build-and-scan: runs-on: callahan:latest ai: review: true explain-failures: true
Each step runs as a shell process under your user. State lives in a single SQLite file. Live logs stream over a WebSocket to the dashboard.
Four providers wired in. Pick one in the settings page; switch any time without a rebuild. No key? Falls back to Ollama.
If Trivy or Semgrep are on the runner, Callahan runs them per build, parses the JSON, and renders findings as expandable cards. If neither is installed, the AI does an LLM source review instead.
Deploy step is literally a shell. kubectl, fly, vercel, terraform — whatever you already know.
HMAC-verified inbound webhooks for GitHub push / pull_request and GitLab Push / MR hooks. Build results post back to the PR as a single markdown comment.
Streaming output, structured events, and a one-click plain-English diagnosis on any red step. No more 800-line stack traces scrolled by hand.
Each is a single LLM call with structured input and output, scoped to one task. Toggle any of them off in the ai: block of your Callahanfile.
Type what your pipeline should do. The Architect writes the YAML — jobs, steps, deploy stages, AI flags — and on save commits it back to the repo using your project's GIT_TOKEN.
Click AI Explain Failure on any failed build. The agent reads the failed job's logs and the Callahanfile, returns a short root cause and a concrete next step.
Reads the diff for the build's commit and writes a structured review with severity, findings, and a fix suggestion. For PR builds, posts the review as a comment on the originating GitHub or GitLab PR.
Runs Trivy and Semgrep when they're on the runner; falls back to AI-only source review when they aren't. Findings are bucketed by severity and posted on PRs alongside the AI review.
Callahan isn't a Jenkins replacement for a 200-engineer org. It's the local-first option for solo devs and small teams who want their CI on their machine, with AI parts that already work.
| Capability | Callahan | Jenkins | GH Actions | Dagger |
|---|---|---|---|---|
| Runs locally | ✔ Yes | ✔ Yes | ~ Self-hosted runners | ✔ Yes |
| Single binary, no DB or queue | ✔ Go + SQLite | ✖ JVM + plugins | ✖ Cloud-managed | ~ Needs Docker |
| AI code review on every build | ✔ Built in | ✖ — | ✖ — | ✖ — |
| NL → pipeline YAML | ✔ Yes | ✖ — | ✖ — | ✖ — |
| Security scanning bundled | ✔ Trivy + Semgrep | ✖ Plugin | ✖ Action | ✖ — |
| Dashboard UI | ✔ Bundled | ~ Dated | ✔ Web | ✖ CLI only |
| Posts results to PRs | ✔ GitHub + GitLab | ~ Plugin | ✔ Native | ✖ — |
| Works without internet | ~ With Ollama | ✔ Yes | ✖ — | ✔ Yes |
| License | MIT | MIT | Proprietary | Apache 2.0 |
macOS, Linux, and WSL2. Docker is the quickest path if you just want to run it.
Go + Next.js source. Everything runs from this one directory.
git clone https://github.com/Callahan-ci/Callahan
For AI features. Skip if you'll use Ollama locally.
echo 'ANTHROPIC_API_KEY=sk-ant-…' >> .env
Backend on :8080, Next.js on :3000. Tail both in one terminal.
cd Callahan && ./start.sh dev
Add a repo, paste a personal access token, click Run.
open http://localhost:3000