Resources
FAQ
Does my code leave my machine?
Not through Modula. The app runs entirely locally: state lives in a local database under ~/.modula/, and there’s no Modula cloud or telemetry. Agents are your own coding agent CLIs running with your credentials, so the only outbound traffic is what those CLIs send to their model providers. To keep even that local, point an OpenCode provider at a local or self-hosted model.
Which coding agents can I use?
Claude Code, Codex, and OpenCode, in any combination. Each agent in your workspace picks its own provider and model, so the Researcher can run on one model while Workers run on another.
Do I need an API key?
No. Modula never talks to model APIs itself. If your coding agent CLI is logged in and works in your terminal, it works in Modula.
Will agents push to my repos or merge anything?
No. Agents commit on isolated branches in worktrees under <project>/.worktrees/, and never push. You review the diff in Modula, push the branch, open the PR, and merge. In direct mode agents commit on the base branch, but still only locally.
How autonomous is it?
Your call. The pipeline runs on its own once you approve a task, but every gate is yours: agents never act on a task you haven’t approved, the Researcher pauses to ask you questions when it’s unsure, you can kick any variant back to rework, and nothing merges without you. Or lean in further and let it run end to end, only showing up to review and merge.
What if two agents work at the same time?
That’s the design. Every variant gets its own branch and worktree per project, so parallel attempts (on one task or across many) never collide with each other or with your own checkout.
Can I pull tasks from my issue tracker?
Yes. Default scan agents mirror issues from Jira, Linear, and GitHub into your task list via MCP servers on your provider; see tracker sync for setup. Synced tasks keep their tracker ID and link, and still wait for your approval before any agent touches them.
Where is everything stored?
Under ~/.modula/: one SQLite database for structured state (tasks, agents, runs, threads), plus per-workspace folders for specs, agent run logs, and the AI Wiki, all as plain markdown and log files you can open with anything.
What does Modula cost to run?
Modula itself is free to use, and the source is public. Agent runs consume your providers’ usage (your Claude subscription, API credits, or local compute). The Agents → Usage page tracks cost and tokens per run, per agent, and per workspace.
What platforms are supported?
macOS (Apple Silicon and Intel), Windows, and Linux. Grab the build for your platform from GitHub releases .
Where do I report bugs or ask questions?
GitHub Issues for bugs and feature requests, GitHub Discussions for everything else.