agav
Agav is an open-source, terminal-native AI coding system/assistant for real repositories. It lets developers inspect code, edit files, run commands and tests, and verify changes from a CLI-first, local-first workflow.
At a Glance
- Individual software developers
- Small engineering teams
- Startups
- Enterprise engineering organizations
- +2 more
AI Tools by agav
(1)Agav
Open Source Terminal AI Coding Agent
Discussions
No discussions yet
Be the first to start a discussion about agav
Latest News
Main branch promoted recent beta changes; repository activity showed the current 0.2.8-era codebase and release workflow.
Agav v0.2.8 released with beta-channel/installer support and ongoing CLI, MCP and sandbox improvements.
Agav highlighted mouse selection, copy, prompt editing and scrolling in its terminal UI.
Agav announced a single native binary for macOS, Linux and Windows, with no Node/npm dependency and no telemetry claim.
Products & Services
An experimental-beta, terminal-native autonomous AI coding assistant distributed as a self-contained binary for macOS, Linux and Windows. It works on real repositories, with interactive and non-interactive/CI modes.
The team/enterprise workflow described by Agav, adding organization-managed policies, SSO, model-gateway routing and repository-aware agent workflows while retaining a local CLI experience.
A companion marketplace repository for installable agents, skills, plugins and extensions for the Agav CLI harness.
A related open-source benchmark for evaluating AI agent harnesses on tool use, long-horizon tasks and cost/token efficiency.
Market Position
Agav positions itself as an open-source, local-first, terminal-native alternative to hosted or more opinionated coding agents such as Codex, Claude Code and OpenCode. Its differentiation is the harness: provider/model choice, repository-aware tools, sandboxing, persistent project memory, scoped subagents, MCP extensibility and explicit cost/verification controls; its own study reports lower model-token cost than Codex in the tested comparison but also a lower pass rate.
Founding Story
Agav was started around the idea that AI coding costs and reliability are strongly influenced by the agent harness, not only the model. Its initial vision was an auditable, Apache-2.0, local-first terminal agent that preserves context, bounds unproductive loops, supports model choice, and keeps repository work and verification together.
Business Model
Revenue Model
The CLI is Apache-2.0 open source and the website says local use is free forever. A hosted/team pricing model is not yet finalized; the pricing page says plans are coming soon. Enterprise customers are invited to contact sales.
Pricing Tiers
Free forever for local use; users supply their own model-provider or Ollama infrastructure costs.
Target Markets
- Individual software developers
- Small engineering teams
- Startups
- Enterprise engineering organizations
- CI/CD and developer-platform teams
- Organizations seeking self-hosted, auditable or provider-agnostic AI coding tooling
- Repository maintenance, bug fixing and focused code changes
- Running tests and verifying changes before human review
- Software development from a terminal without a hosted runtime or mandatory GUI
- Non-interactive automation and CI/CD jobs
- Long-running or multi-step coding tasks using plans, memory and subagents
- Teams that need organization policy, SSO and model-gateway integration