HN Showcase
Back to browse
GitHub Repository

Autonomous AI development engine. Manages multiple repos as a single product — scans codebases, proposes tasks from a product vision, writes code, verifies builds, creates PRs, reviews diffs, and merges — fully wired end to end.

0 starsTypeScript

OpenTangl – Autonomous AI dev engine for multi-repo products

by 8con·Feb 24, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidBig BrainShip It

Cross-repo dependency resolution is clever; but autonomous code agents are a crowded, uncertain category.

Strengths
  • •Solves real friction: understands task dependencies across multiple repos and orders execution correctly
  • •Cost-efficient: runs locally with API calls (~$0.30-0.50 per cycle); no SaaS lock-in
  • •Pragmatic: escalates risky PRs instead of blindly merging; transparent about uncertainty limits
Weaknesses
  • •Autonomous AI agents remain largely experimental; Devin, GitHub Copilot Workspace, and others already own this narrative
  • •No production case studies or evidence of sustained, multi-cycle reliability on real complex projects
Target Audience

Solo developers, small teams, startups wanting to automate routine multi-repo development

Similar To

Devin · GitHub Copilot Workspace · Continue

Post Description

OpenTangl is an open-source TypeScript tool that runs an autonomous development loop. You define a product vision and point it at one or more repos. It proposes tasks, writes code via GPT-4o or Claude, verifies builds, creates PRs, reviews diffs, and merges — in a closed loop.

The part I found most useful: it manages cross-repo dependencies. If a UI feature needs a new API endpoint, it builds and merges the API task first, then builds the frontend. Task ordering via `depends_on` fields.

It runs locally on your machine. No cloud services, no accounts beyond an LLM API key and GitHub CLI. A cycle of 4-5 tasks costs about $0.30-0.50 in API calls.

Built this for my own projects over the past few months. Open-sourced it today. TypeScript, ~17k lines, MIT licensed.

Similar Projects

Developer Tools●●Solid

Prodlint – Find what AI coding tools miss before production

This is a focused, pragmatic tool: 27 rules across security, reliability, performance and AI-quality pick up things TypeScript and ESLint miss (hallucinated imports, phantom-dependency, hardcoded secrets, missing rate limiting). It ships as an npx CLI with JSON output for CI, tests and GitHub Actions — small but practical feature set that makes it trivial to gate AI-generated code in pipelines. I'd like stronger editor/IDE integration and more ecosystem hooks, but for teams relying on LLMs this hits an important pain point cleanly.

Niche GemSolve My Problem
AMARCOVECCHIO99
113mo ago
AI/ML●●Solid

The Rouge is my attempt at an AI product factory

Runs Claude Code with --dangerously-skip-permissions to ship MVPs overnight.

Bold BetBig Brain
gr3gario
311mo ago