I built **aiagentflow** because I kept running into the same problem: AI coding tools either require you to trust your codebase to a third-party cloud, or they give you a single autocomplete-style agent with no real coordination. I wanted something that feels more like a small engineering team: **Architect → Coder → Reviewer → Tester → Fixer → Judge** Each agent runs locally, reads the same specs and guidelines, and works together until the task passes quality gates. The workflow is simple: 1. Run `aiagentflow init` to detect your project, configure agent roles, and import docs. 2. Drop your PRD, API spec, or security guidelines into `.aiagentflow/context/`. 3. Run `aiagentflow run "your task"` and let the pipeline work until the Judge passes it, or you step in. You can mix providers by role. For example, I use Claude for Architect and Reviewer, where reasoning matters most, and a local Ollama model for Coder and Tester to keep costs low. Groq’s free tier works great too. I’d love your feedback: * Which agent roles feel missing? * What would make PRD → task list → batch run more useful? * Would you prefer a VSCode extension or desktop GUI first? Install: `npm install -g @aiagentflow/cli` GitHub: `github.com/aiagentflow/aiagentflow` Docs: `aiagentflow.dev` Happy to answer anything. Thanks for hunting it!

aiagentflow CLI

A local-first AI engineering team in your terminal.

A local-first AI engineering team in your terminal.