How Gentic is structured
Gentic has two parts that work together: the web app you use every day, and the agent worker you deploy once on a server.Web App (gentic.chat)
Create and manage projects, write issues, choose an AI agent, upload attachments, and watch pull requests appear — all from your browser.
Agent Worker CLI
A background process you deploy on any server or VM. It polls Gentic for queued issues, clones the right repo, runs the agent, and pushes a PR.
Key features
Issue Management
Create issues with a title, detailed prompt, agent choice, and file attachments. Move them through Draft → Todo → In Progress → Done.
Claude Code & Codex
Assign each issue to Claude Code or OpenAI Codex. The worker selects the right agent automatically based on the issue setting.
GitHub Integration
Connect your GitHub repositories so the agent worker can clone them, create branches, commit changes, and open pull requests on your behalf.
File Attachments
Attach images, text files, and other documents to an issue. The agent receives them as part of its first prompt, giving it real context to work from.
Issue Dependencies
Block one issue on another to sequence work. Dependent issues won’t be picked up by the worker until their blockers are resolved.
MCP Server
Manage Gentic from Claude Desktop, Cursor, or any other MCP-compatible AI client without opening the web app.
Supported AI agents
Gentic currently supports two AI coding agents. You choose the agent when you create an issue, and you can switch it at any time before the issue moves to Todo.| Agent | Provider | Notes |
|---|---|---|
| Claude Code | Anthropic | Requires Claude credentials in the worker environment. |
| Codex | OpenAI | Requires the Codex CLI installed and authenticated in the worker environment. |
Architecture overview
When you move an issue from Draft to Todo, Gentic queues it for the worker. Your deployed agent worker polls the Gentic API every few seconds, claims the next queued issue, clones the project’s GitHub repository into a fresh working directory, and runs the chosen agent through the Agent Client Protocol (ACP). The agent writes code and opens a pull request, and the issue status updates in real time in your web app. The worker never exposes a public endpoint — it is a private outbound polling process that you run in a trusted environment.Because the agent worker runs in your own environment, it has access to your source code and Git credentials. Run it on a server or VM you control, and treat your
GENTIC_API_KEY as a secret.
