The bottleneck shifted. Your tools didn't.

Claude Code ships features end-to-end, but you're still spending hours figuring out what to build. Still copy-pasting user feedback into Notion. Still writing specs manually that go stale before the agent finishes reading them. Still context-switching between Linear, Slack, and the terminal to coordinate work that an AI could execute in ten minutes — if only someone told it what to do.

The bottleneck in AI-assisted development is no longer code. It's coordination.

Foundry is the orchestration layer between human intent and AI execution. Paste a rough idea. Get agent-executable specs in minutes. Ship it.

What Foundry Does

Idea Decomposition

Paste notes, voice transcripts, or half-baked thoughts. AI breaks them into discrete features.

AI Spec Generation

Chat with a Foundry agent to refine any feature into a full mini-PRD: problem statement, solution, user stories, and acceptance criteria. Ready for handoff to any AI coding agent.

Kanban Lifecycle

Drag features through Idea → Defined → In Progress → Done. Priority ordering, status management, and request tracking — a focused board built for AI-speed development.

Git Integration

Review diffs, generate commit messages, and push — without leaving Foundry. AI-powered code review catches security issues, bugs, and architectural concerns.

PM + Dev Modes

Toggle between non-technical and technical views of the same feature. PMs see outcomes and user stories. Developers see implementation details and acceptance criteria. Same source of truth.

Why Foundry

You have an AI agent that can ship a feature in 10 minutes. You spend 2 hours writing the spec for that 10-minute task.

| Today | With Foundry | | -------------------------------------------------- | ---------------------------------------------------- | | Ideas live in Notion, Slack, scattered notes | Paste raw ideas, AI decomposes into features | | Specs written manually from scratch | Chat-based spec refinement with full project context | | Feedback trapped in emails and interviews | Embeddable widget → auto-extraction → backlog | | Context-switching between Linear, Notion, terminal | One surface: board + specs + git + AI | | Specs written for humans to read | Specs structured for AI agents to execute |

Tools built for human-speed development don't work at AI-speed. Foundry does.

Quick Start

bunx foundry-ai

On first run, Foundry will clone the repository, prompt for your Vercel AI Gateway key, run database migrations, and start the dev server. That's it.

Vercel AI Gateway routes requests to multiple AI providers (Claude, Gemini) through a single API key. You can sign up for free at vercel.com/ai-gateway.

Manual setup

git clone https://github.com/umate/foundry.git
cd foundry
./start.command

Options

bunx foundry-ai --port 3000     # Custom port (default: 5005)
bunx foundry-ai --dir ~/apps    # Clone to a specific directory

Prerequisites

• Bun v1.0+
• A Vercel AI Gateway API key (free tier available)
• That's it. SQLite is embedded. No Docker. No Postgres. No Redis.

Tech Stack

| Layer | Technology | | ------------- | ---------------------------------------------------------- | | Framework | Next.js 16, React 19, App Router | | Styling | Tailwind CSS v4, ShadCN UI (custom retro/industrial theme) | | Database | SQLite via bun:sqlite, Drizzle ORM | | AI | AI SDK v6, Claude Code Agent SDK, Google Gemini | | Runtime | Bun | | Language | TypeScript | | Icons | Phosphor Icons |

Local-first by design. No cloud database. No external services beyond AI. Your data stays on your machine.

Roadmap

Shipped

• [x] Idea → feature decomposition with AI
• [x] AI spec generation via Claude Code agent
• [x] Kanban board with drag-and-drop
• [x] Feedback widget with auto-extraction
• [x] Git integration (diff, commit, push, code review)
• [x] PM and Dev mode toggle

Next

• [ ] Github integration (creating PRs, pulling branches)
• [ ] List view
• [ ] Better context management
• [ ] PRD export and sharing

Contributing

Foundry is open source and contributions are welcome.

1. Fork the repository
2. Create a feature branch (git checkout -b feat/your-feature)
3. Make your changes
4. Run bun run lint to check for errors
5. Submit a pull request

License

MIT

Built with conviction. Foundry was/is used to build Foundry.