Type an idea.
Watch it become a deployed app.
DevFoundry runs an autonomous software factory on your own machine — it designs, codes, tests, and deploys a real application while you watch, using your models (or fully offline).
Why DevFoundry
Cloud AI builders lock you in and take your code. DevFoundry keeps everything on your machine.
Local-first, actually
Your ideas, API keys, and every line of generated code stay on your laptop. No cloud, no telemetry.
Bring your own models
24 providers — Anthropic, OpenAI, Groq, Google, plus keyless OpenCode Zen and fully-local Ollama / LM Studio.
Zero-config
Nothing set up? It auto-detects a running local runtime and just works. Smart rotation & failover across providers.
Agency-grade output
A skills system, house design system, and RAG knowledge base push generated apps to sellable quality.
Free deploys, built in
Netlify, Vercel, Cloudflare Pages, Surge, Hugging Face Spaces — with custom domains — or local Docker/zip.
More than a builder
Deep Research (cited reports), an MCP plugin catalog, a code editor, live preview, and persistent history.
Your models, your machine
Cloud or fully local — and it recommends the best local model for your hardware.
Best local model for your specs
Detects your RAM, chip, and GPU (Metal) and recommends a model that fits — with the exact ollama pull command.
MoE + auto-router
A different expert model per pipeline stage, automatic failover on rate limits, and a one-click rotation built from every free provider you have.
Any gateway
OpenAI-compatible? It works — including self-hosted proxies like FreeLLMAPI.
Inside the app
Real screenshots — nothing mocked.
Get started
# 1. Install the desktop app — download for macOS / Windows / Linux: # https://github.com/KingLabsA/devfoundry/releases/latest # …or build from source: git clone https://github.com/KingLabsA/devfoundry.git cd devfoundry/frontend && npm install && npm run tauri build # …or use the CLI + web app (no desktop build needed): cd devfoundry/backend && python3 -m venv .venv && .venv/bin/pip install -e . .venv/bin/devfoundry serve # web app at http://localhost:9100/app # 2. (Optional) a free local brain — auto-detected: ollama pull qwen2.5-coder:14b # 3. Open the app → pick a Skill → type an idea → Forge It. ✓ spec → design → codebase → refine (build + tests) → deploy
Documentation
Everything is in the repo.