GuideAdvanced

Parallel Coding Agents with Orca

Working with one coding agent at a time has a ceiling: you wait for it to finish, you switch branches, you lose the thread. This guide teaches you to work with **several coding agents at once** (Claude Code, Codex, OpenCode) without them stepping on each other, using **Orca**, an open source desktop app that gives every task its own git worktree, terminals, and browser. You learn it by working through the backlog of **lumen-tasks**, a small task board you build yourself at the start of the guide: first you race three agents on the same bug and keep the best diff; then you hand a task from one agent to another; then you set up a coordinator that splits work across workers with a task DAG and decision points; you verify the UI with the embedded browser and Design Mode; you scale it with automations, GitHub, Linear, stacked PRs, and remote worktrees; and you finish with one feature split across three agents and a coordinator. The guide was written and verified against **Orca v1.4.227** (October 2026). Orca ships new versions almost daily: every command in the guide comes from its official documentation, and you learn how to read the guide that ships with your own version.

64
lessons
8
modules
English · Spanish
available in
Free
access
NIEVA

Outcomes

What you'll be able to do

  • Decide when parallelizing agent work pays off and when it doesn't, counting the cost in tokens and in your own review
  • Explain what a git worktree isolates and what it doesn't (ports, secrets, network), and why a worktree is not a sandbox
  • Install Orca, connect your agents, and create worktrees with different agents on the same repo
  • Run the same task in three agents, compare their diffs with a rubric, and keep the winner
  • Drive Orca from the terminal and from Claude Code with its CLI: read, wait on, and send to agent terminals
  • Hand a task from one agent to another with a handoff brief that keeps the context
  • Orchestrate a coordinator and its workers with tasks, dependencies, questions, escalations, and human decision points
  • Verify UI changes with each worktree's browser, Design Mode, and browser automation from the CLI
  • Scale with scheduled automations, GitHub reviews, Linear, stacked PRs, and remote worktrees over SSH
  • Control the risks of working in parallel: merge conflicts, tokens, secrets, permissions, and human review

Before you start

What you need to bring

It's for you if...

  • Developers who already use a coding agent every day and feel they wait more than they work
  • Tech leads who want to try several agents on the same task before deciding which one to work with
  • Anyone who wants to go from "launch agents and hope" to coordinating them with file ownership, acceptance criteria, and review

Requirements and materials

  • Everyday git: commit, branch, merge, and resolving a simple conflict
  • Having used at least one coding agent in the terminal (Claude Code, Codex, or OpenCode)
  • Python 3.12 or later installed (the sample repo uses only the standard library)
  • Access to one or more coding agents. Three different ones is ideal; with only one, you launch it three times and the guide still works
  • For module 7: a GitHub account where you can create a private repository

Content

The syllabus, module by module

Open any of them to see its lessons.

Common questions

What people usually ask

Start whenever you like

Reviews

What students say

These reviews are from enrolled students who completed at least 50% of the course. We moderate reviews only on content grounds (spam, offensive language, personal data), never for being critical or negative.

No approved reviews yet.

Be the first to share your experience!