Open source · MIT · v1.1.0

The AI coding agent that runs free, on your own machine.

Watch a local model reason on screen, then use real tools to do the work — no cloud, no API bill, ~$0. Or bring any of 15 providers with automatic failover. From your terminal, a desktop app, your phone, or a 24/7 service.

⭐ Star on GitHub ✅ See it work ($0) Quick start →

🆓 Local Ollama · $0 🔌 15 providers 🧪 27K+ tests ✅ reproducible proof

Real gpt-5.5 chat streaming in the Cowork desktop app for $0

Real gpt-5.5 in the desktop app — the answer streams in, cost $0.0000.

One engine. Everywhere you work.

🆓

Free & local-first

Runs entirely on local Ollama ($0), any of 15 providers with auto-failover, or a flat-fee ChatGPT Plus/Pro login — no per-token metering.

🧠

Reasoning you can watch

Local models think step-by-step on screen, then call tools to act — edit files, run commands, search the web, verify results.

🛠️

~110 tools, MCP & skills

Edit, shell, web search, browser, PDFs/Office, a skills marketplace, and MCP connectors — extend it to your stack.

🤝

Multi-AI fleet

Peers on your network observe each other live and call each other's models & read-only tools — one machine's spare GPU becomes everyone's.

🤖

24/7 autonomy

A background service claims & runs tasks free-first on local models — TTL leases, DAG deps, a workers→verifier→synthesizer swarm.

👁️

Personal companion

Optional: bidirectional voice, opt-in camera/presence, persistent memory, and 20+ messaging channels.

The Cowork desktop cockpit

Chat, tools, traces, workflows, settings, permissions, models — against the same core engine as the CLI.

Cowork desktop home with quick-action cards and the panel rail — Desktop home — quick-action cards, gradient hero, panel rail

Onboarding to a real first reply from local Ollama, $0 — Zero → first chat: a **real reply from local Ollama**, `$0.0000`

Fleet and autonomy dashboard — Fleet dispatch · tool-permission posture · Hermes toolsets

In your pocket — over Telegram

The same agent as the terminal, reachable from your phone as a messaging bot. The prompt & tools scale to each question — instant for plain chat, loading tools on demand (the same pattern as Codex / Claude). You can also talk to it: a voice note in → speech-to-text (faster-whisper) → reply by voice (Piper TTS), all local / $0 (needs the local voice engines; degrades to text otherwise). Real, unedited captures (the bot is named “Lisa” here).

Telegram: instant greeting, the time, and tomorrow's live weather in Paris via web search — Chat + **live tools on demand** — “what time is it?” and “tomorrow’s weather?” pull the time & `web_search` tools, only when asked

Telegram: the agent confirms it can read and inspect its own source code via view_file — Reads its own source (or any file) via `view_file` — then introduces its recursive self-improvement →

Telegram: the agent explains its recursive self-improvement — Manus-inspired lessons in RULE / PATTERN / CONTEXT categories, stored in .codebuddy/lessons.md — **Improves itself across sessions** — the Manus-inspired `lessons_*` loop extracts **RULE / PATTERN / CONTEXT** lessons into `.codebuddy/lessons.md` (project + global). Matches its real source.

What's shipped — not roadmap

$0 local coding agent — a local model reasons, then calls tools to do real work.
ChatGPT Plus/Pro → gpt-5.5 at $0 — buddy login, flat-fee, no API key.
Goal loops (Ralph loop) — a judge re-checks completion each turn and auto-continues until done.
Multi-AI fleet — peer.chat / peer.tool.invoke across your network.
15 providers with failover & circuit breakers; ~110 tools, MCP, skills.
27K+ Vitest tests wired to CI.

Honest about scope — see the parity doc for exactly what's full vs. in-progress.

Install in seconds

# Free & local: point at a local Ollama, $0
npm install -g @phuetz/code-buddy
export CODEBUDDY_PROVIDER=ollama
buddy

# …or log in with your ChatGPT Plus/Pro subscription (no API key)
buddy login && buddy

Node ≥ 18 for the CLI · Node ≥ 22 for the desktop app. Run buddy doctor to check your environment.

⭐ Star on GitHub Documentation