GitAgent
GitAgent is a git-native, multimodal AI-agent harness and the reference implementation of the OpenGAP standard. It puts an agent's identity, memory, rules, tools, skills, knowledge, and compliance artifacts in ordinary version-controlled repository files so agents can be reviewed, forked, branched, rolled back, and run across model providers and runtimes.
At a Glance
- Developers and engineering teams
- Enterprise AI and platform teams
- Financial services and other regulated industries
- Organizations requiring on-premises, local, or air-gapped deployment
- +1 more
AI Tools by GitAgent
(1)
OpenGAP
Git Native AI Agent Standard
Discussions
No discussions yet
Be the first to start a discussion about GitAgent
Latest News
GitAgent repository merged the Ira Rust port and Lyzr Edgespace desktop application work.
MCP client support was added to GitAgent.
GitAgent 2.0 split voice mode into the optional @open-gitagent/voice package and slimmed the core package.
The reference implementation and CLI were rebranded from gapman to OpenGAP, while gitagent remained a compatibility alias.
Products & Services
Open-source, self-hosted, multi-model coding-agent runtime that runs against local or remote Git repositories; sessions become branches and commits, and learning or memory updates can be proposed through pull requests.
TypeScript SDK for embedding GitAgent in Node.js applications, including query streaming, local-directory or remote-repository operation, tools, workflows, scheduled jobs, and multi-agent handoffs.
Framework-agnostic, git-native open standard for defining agents as repository files; the OpenGAP CLI is its reference implementation and supports adapters for multiple agent runtimes.
Optional voice-mode and web-UI package split from GitAgent core in version 2.0.
Market Position
GitAgent positions itself as the portable, Git-native alternative to framework- or vendor-locked agent systems: the agent definition remains in a repository and can be exported or adapted across Claude Code, OpenAI Agents SDK, CrewAI, Lyzr, OpenClaw, and other runtimes. Its differentiators are reviewable/reversible state, PR-based human approval, self-hosting, and compliance-oriented auditability. Its own documentation contrasts this approach with general-purpose assistants such as OpenClaw and with conventional framework-specific agent runtimes.
Leadership
Founders
Shreyas Kapale
Creator and maintainer of GitAgent; a Founding Architect/Chief Architect and founding engineer at Lyzr Research Labs, with work on AI-agent infrastructure and reliability.
Khush Patel
Co-maintainer of GitAgent and Founding Architect/founding engineer at Lyzr, focused on agent infrastructure, responsible AI, evaluations, safety, and reliability.
Executive Team
Shreyas Kapale
Creator and Maintainer of GitAgent; Founding Architect at Lyzr
Lyzr founding engineer/architect focused on AI-agent infrastructure; creator of GitAgent and author of the GitAgent patterns article.
Khush Patel
Maintainer; Founding Architect at Lyzr
Lyzr founding engineer/architect working on agent infrastructure, responsible AI, evaluations, safety, and reliability.
Founding Story
The project was started at Lyzr Research Labs in response to three problems in the agent ecosystem: agents lacked a canonical source of truth, memory and behavior were trapped in framework-specific formats, and prompt/configuration changes lacked branching, diffing, review, and rollback. The initial vision was to make the repository itself the portable agent, using Git's existing collaboration and version-control workflows.
Business Model
Revenue Model
Open-source MIT-licensed software distributed as a self-hosted CLI, npm packages, and an embeddable SDK; the public materials describe free/open-source usage rather than a hosted subscription or usage-metered model.
Pricing Tiers
MIT-licensed GitAgent/OpenGAP code and npm packages; users provide their own model-provider credentials and infrastructure.
Target Markets
- Developers and engineering teams
- Enterprise AI and platform teams
- Financial services and other regulated industries
- Organizations requiring on-premises, local, or air-gapped deployment
- Teams building multi-model or framework-portable agents
- Autonomous coding and project agents
- Enterprise AI assistants in banking, insurance, healthcare, and other regulated environments
- Developer copilots that learn a codebase and submit pull requests
- Multi-model and multi-framework agent workflows
- Team-shared agents managed with Git branching and code review
- Knowledge bases, persistent memory, and auditable agent operations