Index / Gitagent · updated Oct 2, 2026
Gitagent
open-gitagent/gitclaw · healthy · rank 28 of 43 by stars
A git-native agent framework where identity, memory, rules, and tools are all version-controlled files — fork an agent like a repo. Recently hardened with security fixes and a Rust rewrite ('ira') powering a Tauri desktop app.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 78 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 57 files examined
3 providers · runs locally · custom endpoint
Pinned models
gpt-4o-mini released 2024-07-18 — 2 years old (from the public model catalogue)
Pin last edited 44 days ago
Evidence
Decision
Why choose Gitagent over OpenClaw?
Why choose this
- Git-native agent definition enables diffing, branching, and full audit history
- Recent security hardening (RCE fix, race fixes, leak-free cleanup) shows strong engineering rigor
- Rust port with Tauri desktop app offers lower footprint and local model management
Tradeoffs
- Far smaller community and zero Reddit presence
- No multi-channel messaging integrations comparable to OpenClaw
- Plugin/skill ecosystem is young and unproven
Best fit
- Developers who want agent config and memory as auditable git history
- Teams forking/branching agent personalities and rules
- Users wanting a desktop app with local model support via Ollama
Avoid if
- You need a large, proven community and plugin marketplace
- You want a messaging-first assistant (WhatsApp/Telegram/Slack channels)
- You need battle-tested multi-user deployment
Evidence is solid on architecture and recent engineering activity from README and commits, but community sentiment is essentially unmeasurable (0 Reddit matches, generic web results). Security and performance estimates are inferred from commit content rather than independent audits.
AI layer reviewed Sep 21, 2026 · how this is written
Star activity
711 stars today
Overview
Gitagent reimagines the AI agent as a git repository: agent.yaml holds model and runtime config, SOUL.md defines personality, RULES.md encodes behavioral constraints, and memory/, tools/, skills/, and hooks/ are all version-controlled files. This 'agents as repos' model lets you fork an agent, branch a personality, and git log its memory — a genuinely distinctive architectural stance compared to OpenClaw's platform-style approach.
The project is in active, high-quality development. Recent commits show two hardening batches closing a load-time RCE via shell-interpolated git clone, fixing tool concurrency races, plugging e2b sandbox and credential leaks, and adding proper abort/steer semantics. It also shipped MCP client support (v2.1.0), session-scoped telemetry tracing, and a growing unit test suite. Distribution is via npm with a slim core package explicitly designed to pass supply-chain scanners, plus an optional voice/web UI package.
Most notably, the team is porting the engine to Rust ('ira') bundled with a Tauri desktop app (Lyzr Edgespace branding) featuring an embedded web UI, Memberstack auth, Ollama-based local model management with self-healing serve, and a first-run onboarding wizard. This positions Gitagent as a local-first, desktop-friendly alternative to OpenClaw — though its community, plugin ecosystem, and messaging integrations remain far smaller.