Hey Product Hunt 👋 We’re moving into a world where more and more of software development can be defined with prompts. Agents can write code, reason about tasks, use tools, review changes, and work across an entire repository. That made me wonder: **why should coding agents themselves be locked inside a specific company's infrastructure, models, and token ecosystem?** I wanted something where I could be in charge of the agent. So I built **Jr-Arch**. With Jr-Arch, you can define your own agents, personas, rules, and guardrails as files inside your repository. You choose the model, bring your own API key, or run a local model. The agent system becomes part of your codebase rather than something completely controlled by a platform. It's built around the **OpenGAP protocol**, supports multiple model providers and OpenAI-compatible endpoints, and has no telemetry or runtime dependencies. The idea is simple: **Your repo. Your agents. Your models. Your rules.** I built Jr-Arch because I wanted to see what coding agents would look like if the developer, not the platform, was in charge of the agent layer. It's open source, and I'd genuinely love feedback from people building with AI agents. What would you want to control if your coding agent was completely yours? Try it: `npx jr-arch`

Jr-Arch

Coding agent in your repo. Your models, agents & rules.

Coding agent in your repo. Your models, agents & rules.