<div align="center">

Agav

Terminal-native autonomous agent for real repositories and projects

<p align="center"> <a href="https://docs.agav.dev"><strong>Documentation: https://docs.agav.dev</strong></a> </p> <p> <img alt="Version" src="https://img.shields.io/github/package-json/v/prapaa-ai/agav?style=for-the-badge&amp;label=version&amp;color=111"> <img alt="License" src="https://img.shields.io/badge/license-Apache%202.0-111?style=for-the-badge"> <a href="https://github.com/harbor-framework/terminal-bench-2-1/pull/225"><img alt="Terminal-Bench 2.1" src="https://img.shields.io/badge/Terminal--Bench_2.1-84.7%25_%7C_top_of_the_board-111?style=for-the-badge"></a> <a href="https://swe-bench-live.github.io"><img alt="SWE-bench-Live Lite" src="https://img.shields.io/badge/SWE--bench--Live_Lite-62.0%25_%7C_%232_on_the_board-111?style=for-the-badge"></a> </p> </div> <div align="center"> <img src="https://www.agav.dev/preview.gif" alt="Agav preview" width="100%" style="border-radius:16px;" /> </div>

⚠️ Experimental Beta — Agav is under active development. Expect breaking changes, rough edges, and rapid iteration. Feedback and bug reports are welcome!

Install

No Node.js required. One command, self-contained binary:

curl -fsSL https://agav.dev/install.sh | bash

Windows PowerShell:

irm https://www.agav.dev/install.ps1 | iex

Windows Command Prompt:

curl -fsSL https://agav.dev/install.cmd -o install.cmd
install.cmd
del install.cmd

Download binaries or pre-releases from Releases. See docs.agav.dev/getting-started/installation for detailed installation steps.

Run

agav

Pick a provider and model, or use defaults:

agav --provider openai --model gpt-4o
agav --provider openrouter --model openrouter/auto
agav --provider ollama --model llama3.2

Non-interactive for scripts and CI:

agav run "review the code in src/"
agav -P "what does this project do?"
cat error.log | agav -P "explain this error"

Keep it current:

agav update

Why Agav

Today, the most powerful AI tools are controlled by a handful of companies. They choose which models you use, where your data goes, and how much you pay. If they change the rules tomorrow, you have no choice but to follow.

We’re building the alternative, an AI that works for you, runs on your terms, and belongs to no one.

You own it. Apache 2.0 open source. No lock-in, no hidden telemetry, auditable code.

You choose the brain. 7 providers — Anthropic, OpenAI, OpenRouter, NVIDIA NIM, Gemini, Vertex AI, Ollama. Switch models mid-session with /model, run local with Ollama, or route via OpenRouter.

You control the data. Encrypted API keys, local-first option, per-tool permissions, and sandboxed commands on macOS/Linux.

Autonomous and verifiable. Plans are created and tracked automatically, human-in-the-loop for destructive actions, and non-interactive agav run for CI with JSON Schema output.

Proof

• Terminal-Bench 2.1 — 84.7% 377/445 trials ±0.84%, top of public board.(PR under review)
• SWE-bench-Live Lite — 62.0% 186/300, #2 on the board.

Highlights

• Providers — switch mid-session with /model. Fast /fast and deep /deep shortcuts.
• Non-interactive mode — agav run and agav --print with per-tool permissions and optional JSON Schema output.
• Sessions that survive — resume, branch, name, search and export. /compact reclaims context without losing plans.
• Skills & Agents — reusable instruction bundles and installable service agents from a marketplace. Delegate scoped work to fresh-context subagents.
• Memory — cross-session memories per project: user preferences, feedback, project decisions, references.
• Repository-aware editing — LSP queries, notebook support, test running, /undo, and planning UI.

Quick links

• Quick start
• CLI reference
• Tools, skills & agents
• Workflows & automation
• Changelog

Community

Email: contact@agav.dev
Discord: discord.gg/6u3m2JN6k

License

Apache 2.0

---

Documentation source lives in docs/. To run locally: cd docs && npm install && npm run dev.