session-orchestrator
Your task becomes tested work, and the next session picks up where the last one stopped.
For developers · Open source, MIT. · session-orchestrator.com (opens another website)
What you do with it
You want a download button on your customer list, so you tell your coding agent. It first looks at where the project stands and suggests a plan. Once you agree, several agents build in rounds, and after each round the tests run to see whether everything still works. If you close your laptop in the evening, the next session knows the next morning what is done and what is open.
- Who it's for
- For people who build their own project with Claude Code or another coding agent and would rather not explain everything again each session.
- Where it stands
- Open source under the MIT licence, available on npm. It brings no language model of its own; your coding agent still writes the code.
Step by step
Start with /session
You describe the task, and the agent reads where the project stands and suggests what to work on. Nothing changes until you agree.
Let /go do the building
Several agents work in rounds, each on its own files. The tests run after every round, and anything that fails goes into the next one.
Wrap up with /close
The agent compares what was planned with what got done and makes a commit. Anything still open becomes a task.
Pick up the next day
The next session reads what was written down and offers to carry on from there. It is kept as plain text in your project, no account needed.
How it works
- No language model
- in the tool. Your coding agent writes the code.
- The code
- gives the agent its workflow: plan, work in rounds with tests after each one, wrap up. The state is kept as plain text in your repository.
Why it exists
It began with a Notion page full of prompts. Before every session I looked for the right line, copied it and explained once again how I like to work.
The proof
Commits in every month since April 2026.
Further reading
Part of what you get
This product is an example of: