FAQ

Does Jimothy send my code anywhere?

Jimothy itself has no server. It talks only to the services you configure: your issue trackers, LLM APIs you set up as harnesses, notification webhooks, and Tailscale if you enable remote access. The agents you run (Claude Code, Codex, …) send code to their own model providers, as they would if you ran them yourself.

Do I need an issue tracker?

No. Start runs by hand with New run (⌘N), on a schedule, or from any system that can POST JSON to a webhook trigger.

Do I need to install an agent to try it?

No. The Demo pipeline uses the built-in Simulator, which needs nothing. Real work needs at least one CLI harness or an API key.

Which harness should I use?

Use a coding CLI (Claude Code, Codex, Gemini CLI, Cursor Agent, Copilot, Aider, OpenCode, Pi, Amp, Goose) for any step that reads or changes code. Use an API harness for text-only steps like triage or summaries. Mixing them per step is the point: the Feature template plans with Claude Code on Opus, builds with Sonnet, and reviews with Codex. See Harnesses overview.

How much does a run cost?

Whatever the harnesses charge. Jimothy records cost when a harness reports it (Claude Code, Pi) and tokens when it reports those (Claude Code, Codex, Pi, the API harnesses). Per-step and per-run figures are on the run page; the Dashboard shows 7-day spend. Harnesses that only report tokens contribute $0 to the cost figure.

Can several runs happen at once?

Yes. Settings → Max concurrent runs (default 3) caps runs across all pipelines, and each pipeline has its own Max concurrent runs (0 = no extra limit). Use Git worktree workspaces so parallel runs don’t touch the same files. Steps inside a run also run in parallel when they share a dependency.

What happens if I close the window?

Jimothy keeps running in the system tray: triggers keep polling and runs continue. Use Quit in the tray menu to exit. Runs in progress when you quit are marked interrupted and can be retried.

What happens to the branch and workspace after a run?

They’re kept by default, so you can inspect or retry. Turn on Delete a run’s workspace after it succeeds in Settings to remove worktrees and folders after successful runs (branches are kept). Deleting a run never deletes its workspace.

Can I use it from my phone?

Yes, over Tailscale. See Remote access and Approvals from your phone.

Can Jimothy open the pull request?

Yes, through a shell step. The built-in templates end with git push -u origin HEAD && gh pr create …, which needs the GitHub CLI signed in. PR links in step output are picked up automatically and shown on the run and in notifications and write-back comments.

Does it work with GitLab or Bitbucket?

Workspaces use plain git, so any remote works, and GitLab merge request and Bitbucket pull request links are recognised in step output. There are no GitLab or Bitbucket issue triggers; use a webhook trigger to start runs from them, and a CLI such as glab in a shell step to open merge requests.

Can I run local models?

Yes. Use the OpenAI-compatible API harness with Ollama, LM Studio or vLLM for text steps, and a CLI like Aider pointed at your local model for code-editing steps. See Local models with Ollama.

Where is my data, and how do I back it up?

In the data folder shown at the top of Settings (~/Library/Application Support/Jimothy/factory on macOS). Copy the folder to back it up; secrets are tied to the machine’s keychain. See Data & security.

How do I reset everything?

Quit Jimothy and delete (or rename) the data folder. On next launch you get a fresh install with the Demo pipeline.

Can two people share one Jimothy?

Jimothy is a single-user desktop app. Remote access lets you reach it from your own devices; anyone with the sign-in link gets full control, and approvals are recorded under Settings → Your name. For a team, each person runs their own Jimothy with their own triggers (use distinct labels or assignees so they don’t pick up the same issues).