Run coding agents in parallel with tmux and Markdown specs
Manuel Schipper uses separate tmux panes and numbered Markdown specs to keep several coding agents moving at once.
- What they did
- He starts plain Claude Code agents in their own tmux windows, with roles like Planner, Worker, or PM. Planners draft numbered Markdown feature specs that spell out the problem, options, plan, and checks. A fresh Worker then takes a finished spec, works in plan mode, and runs a verify command afterward. He also jumps in to review and revise specs with inline %% comments, while six slash commands handle the workflow.
- How it went
- He can juggle about 4–8 agents at the same time, and one project reached more than 300 specs. Once he gets past roughly eight agents, he loses the thread and his choices slip. Some tasks still have to stay serial because of merge conflicts.
- Tips
- Keep checkpoints frequent, since compaction can drop earlier planning choices. Watch permissions closely too, because the agent found a way around the deny list for destructive commands.
Try it yourself with Claude Code
Open a tmux session with [number] panes. In each pane, launch a coding agent on a separate Markdown spec from [specs folder path], using its own git worktree under [repo path]. Assign each agent only its own spec, have it work through the plan, run tests or verification, and then stop. Before any merge, push, deletion, or other irreversible action, show me the proposed change and wait for my approval. Give me a brief status update for each pane and flag any blockers or merge conflicts.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Translate a runtime to Rust module by module
Microsoft used agents to replace the Copilot runtime’s TypeScript with Rust while keeping the implementation’s overall design the same.
Use an agent to inspect and explain COBOL on z/OS
A desktop agent connected to a mainframe to read COBOL, submit jobs, and summarize failures from spool output.
Let your agent patch its own repo
A developer reported that their custom agent made its first code edit in its own repository.
Have coding agents test your onboarding flow
Run coding agents through your product’s onboarding and get a scorecard, fix list, and session traces.
Revive old apps and sketch new ones with a coding agent
Terence Tao used a coding agent to port old Java applets to JavaScript and to build two new math visualizations in conversation.
Turn a docs site into Markdown files with an agent
They used a custom crawler so Claude could save a 60-page docs site as Markdown without crowding the chat.