v1.0.0 — open source · MIT

CI/CD that
thinks for itself.

AI-native, serverless pipelines that run entirely on your machine. No cloud bill. No agent clusters. No plugin graveyard.

Get Started Free → Star on GitHub
bash — ~/projects
$curl -fsSL https://callahanci.com | sh
✔ Callahan CI v1.0.0
✔ Detected: macOS arm64 · Docker 26.1
✔ Backend started on :8080 (329ms)
✔ Dashboard started on :3000 (1.2s)
→ http://localhost:3000
<2s
cold pipeline start
0
plugins required
10+
LLM providers supported
50
concurrent jobs on a laptop
MIT
free forever, open source

// capabilities

Everything your pipeline needs.
Nothing you have to install.

Jenkins took 20 plugins to do what Callahan does out of the box. We counted.

Pipeline
Human-readable pipelines
GitHub Actions–compatible YAML with an AI extension. Or just describe it in English — Callahan writes the YAML for you.
name: production on: [push, pull_request] jobs: build-and-scan: runs-on: callahan:latest ai: review: true explain-failures: true
Execution Engine
Serverless. Ephemeral. Stupid fast.
Every job gets a fresh container. Auto-caches npm, pip, cargo, Maven, Go. Parallel execution limited only by your CPU cores.
<2s
cold start
300ms
warm cache
50
parallel jobs
0$
cloud spend
AI Layer
Any LLM. Swap anytime.
Bring your own key or run fully offline with Ollama. Switch per project, per step.
claude gpt-4o gemini grok mistral groq ollama + local
Security
Scans built-in, not bolted on.
Trivy, Semgrep, gitleaks, OWASP, SBOM. All pre-configured.
✔ trivy image scan — 0 critical
✔ semgrep — no issues found
✔ gitleaks — no secrets
✔ sbom generated
✦ AI triage: all clear
Deploy
Every deploy target covered.
K8s, Helm, Terraform, Lambda, Fly.io, Vercel, Netlify, Railway, Render — zero config.
kubernetes fly.io vercel lambda helm terraform
Live Logs
Real-time build output. AI explain on every failure.
Stream logs live. One click to get a plain-English explanation of why a step failed.
▶ step[1] — install dependencies
✔ npm ci (12.3s, cached)
▶ step[2] — run tests
✔ 142 passed, 0 failed (8.1s)
▶ step[3] — build
✖ Error: ENOMEM — out of memory
✦ AI: increase Node --max-old-space-size
GitOps
Full PR workflow out of the box.
AI review comments on every PR. Status checks that block merge. Auto-merge on green. Works with GitHub, GitLab, Bitbucket, Gitea.
✦ AI Review: possible N+1 query on line 84
✔ Status check: build passed
✔ Status check: security passed
✔ Auto-merged: all checks green

// ai agents

Four agents.
Always running.

Not chatbots. Not integrations. Agents woven into the platform that take real action on your code and pipelines.

AGENT 01 / PIPELINE ARCHITECT
Describe it. Get a pipeline.
Type what you want in plain English. The Architect writes your entire Callahanfile.yaml — build steps, test runners, security scans, deploy targets — and commits it to your repo.
Prompt
"Build my Next.js app, run Playwright tests,
scan with Trivy, deploy to Fly.io on green PRs"
AGENT 02 / BUILD DEBUGGER
Failing build? Ask why.
When a step fails, the Debugger reads your full log context, identifies the root cause, and gives you an exact fix — not a generic suggestion. Available in a live chat panel inside any build.
Diagnosis
"Step 3 failed: the container exposes port 8080
but your test expects 3000. Fix: add PORT=8080 env."
AGENT 03 / CODE REVIEWER
AI review on every PR. Automatically.
Every pull request gets a structured AI review posted as a comment — bugs, performance issues, security concerns. Configurable strictness. Works with GitHub, GitLab, Bitbucket, and Gitea.
Review comment
"Line 47: user input passed directly to SQL query.
Possible injection. Use parameterised statements."
AGENT 04 / SECURITY ANALYST
CVEs in plain English. Fixes included.
Trivy and Semgrep findings are parsed, prioritised, and explained by the Analyst. Critical issues block the pipeline. Common CVEs get auto-generated fix suggestions with code.
Finding
"CVE-2024-1234 (Critical) in [email protected].
Upgrade to 4.17.21. Here is the patch diff."

// comparison

How it stacks up.

Everything else either needs the cloud, needs a PhD to configure, or needs a credit card. Usually all three.

Feature Callahan CI Jenkins GitHub Actions Dagger
Self-hosted / local-first ✔ Yes ✔ Yes ✖ Cloud ✔ Yes
Zero config for most projects ✔ Yes ✖ No ~ Partial ✖ No
Built-in AI agents ✔ Yes ✖ No ✖ No ✖ No
Natural language pipeline creation ✔ Yes ✖ No ✖ No ✖ No
Built-in security scanning ✔ Trivy, Semgrep, gitleaks ✖ Plugin only ✖ Action only ✖ No
Dashboard UI ✔ Built-in ~ Basic ✔ Cloud UI ✖ No
Cold start time ✔ < 2s ~ 15–60s ~ 20–45s ✔ Fast
Works offline ✔ Fully ✔ Yes ✖ No ✔ Yes
License MIT MIT Proprietary Apache 2.0

// get started

Running in under 5 minutes.

Mac, Linux, and Windows WSL2. Requires Docker Desktop or Go 1.22+.

STEP 01
Install
One command downloads the binary and sets everything up.
curl -fsSL https://callahanci.com | sh
STEP 02
Start
Auto-detects Docker or falls back to native Go mode.
cd callahan && ./start.sh
STEP 03
Open
Connect your first repo and run a build.
open http://localhost:3000
STEP 04 / OPTIONAL
Add AI
Drop in an API key to unlock all AI agents. Or use Ollama for offline mode.
ANTHROPIC_API_KEY=sk-ant-... ./start.sh

Ship with confidence.
Start today.

Free, open source, and runs entirely on your machine. No credit card. No cloud. No catch.

Get Started Free → View Source
© 2025 Callahan CI Contributors Privacy  ·  Terms  ·  GitHub