Frequently Asked Questions · Foundry docs

Frequently Asked QuestionsFAQ

Information

Still unsure which edition fits? Production-Grade comes with a year of updates and support, so ask us before you decide.

See editions

Does Foundry cost money?Foundry is free and open-source (MIT). You bring your own keys (BYOK) for providers (Anthropic, Google, etc.). You pay only for the inference you use.Does it support local LLMs?Yes. Configure Ollama (e.g. phi4-mini) for the helper/validation agent. That reduces cloud cost and latency for verification steps.Can I run it without Docker?DragonflyDB is required for state and queue. Docker (or Docker Compose) is the standard way to run it. You could run DragonflyDB natively if you prefer.What providers are supported?Cursor (CLI), Gemini, GitHub Copilot (CLI), Claude Code, Codex. Provider strategy and priority are configurable via env (e.g. PROVIDER_STRATEGY, CLI_PROVIDER_PRIORITY).Can I use pnpm?Yes. Use pnpm run cli -- instead of npm run cli -- for all commands. The repo recommends pnpm for the workspace.Does Foundry have a UI?Yes. Foundry includes a full React + Express dashboard with project management, task CRUD, log viewer, settings page, and chat visualizer. Launch it via npm run setup and visit http://localhost:5173.Can tasks run in parallel?Yes. V2 introduces a parallel scheduler with a dual-queue system (ready/waiting). Tasks with depends_on wait until dependencies complete, then auto-promote. Up to 3 workers run concurrently, each isolated via git worktree.What about environmental impact?Foundry's hybrid control plane reduces unnecessary inference calls. Local pre-processing handles routine validation, session reuse prevents context re-injection, and AI is only engaged as a last resort.