Learn
The ideas behind running coding agents, explained from the start and without assuming you have done it before. Each page teaches the concept first. Fleet shows up at the end, as one way of applying it.
Foundations
- What a git worktree isA second working directory attached to the same repository, on a different branch, without a second clone. The plumbing, the commands, and when it beats stashing.
- Why parallel agents need isolationWhat actually goes wrong when two agents share a directory, and the four things you have to isolate: files, ports, processes and history.
- What an agent harness isThe model is one part. The harness is the loop, the tools, the permission gate and the transcript around it, and it decides most of what the experience feels like.
Working with agents
- CLAUDE.md explainedA file in your repository that becomes part of every agent's instructions. What to put in it, what to leave out, and why most of them are too long.
- MCP explainedThe Model Context Protocol is a plug standard for giving an agent tools. What it is, what it is not, and the security question you should ask before adding one.
- Permission models for coding agentsApproving everything trains you to approve everything. The useful question is not which tools an agent may use, it is how far it may reach.
Shipping
- Reviewing a diff you did not writeAgent-written code fails differently from human-written code. A reading order, the specific things to distrust, and how to give feedback that produces one good round instead of three bad ones.
- One agent or a squadParallelism only pays when the work has seams. How to tell whether yours does, and what the coordination actually costs.