Multi-step pipelines
Run steps in parallel or in sequence, with retries, timeouts, conditional runIf steps, and continueOnError.
Pipelines of AI agents, shell commands, and human approvals, running on your machine.
Jimothy is a local-first desktop app that runs multi-step pipelines combining AI coding agents, shell commands, and approval gates. Kick them off from an issue, a schedule, or a webhook, and your code stays on your machine until you approve shipping it.
Run steps in parallel or in sequence, with retries, timeouts, conditional runIf steps, and continueOnError.
Mix AI harnesses per step, including Claude Code, OpenAI Codex, Gemini, Cursor, and GitHub Copilot.
Pipelines pause for review at the steps that matter, so nothing ships without a person signing off.
Every run gets its own worktree, so parallel runs never step on each other.
Start from Linear, Jira, or GitHub issues, schedules, or webhooks, and post comments and status updates back.
Watch each step run with a live visualization and streaming logs.
Get notified on desktop, Slack, Discord, Teams, and more when a run needs you or finishes.
Your data lives in plain JSON files, and secrets are encrypted with your OS keychain.
Code stays on your machine until you explicitly approve shipping it.
Development teams automating review, maintenance, and delivery, plus support and product teams automating work like support replies, lead research, and weekly updates.
Get the app, read the docs, and see everything it can do at jimothy.dev.
Also from EchelonOne: Echelon Inbox · Need help with AI at your company? Work with Brandon