This repository was created from the Machine Foundry template. It is pre-configured to work with the Machine Foundry orchestrator out of the box.
Machine Foundry is an orchestrator that watches this repository's GitHub Projects board and automatically dispatches AI coding agents to work on issues that reach the Ready column.
The default board shape ships ready to use:
| Column | State | Meaning |
|---|---|---|
| Backlog | backlog |
Not yet ready; parked for later |
| Ready | ready |
Agent will be dispatched automatically |
| In Progress | in_progress |
Agent is actively working |
| In Review | waiting |
Agent has finished; awaiting review |
| Human Required | waiting |
Blocked on human input |
| Done | done |
Work complete |
When Machine Foundry starts, it shapes the GitHub Projects board to match these columns automatically.
Edit WORKFLOW.md to:
- Rename or recolour columns — change
nameandcolorfreely; thestatevalue controls behaviour, not the name. - Add columns — add entries with any of the five states:
backlog,ready,in_progress,waiting,done. - Change the agent prompt — replace the Markdown body below the
---front matter separator with whatever instructions suit your project.
Set the following environment variables in your Machine Foundry configuration:
GITHUB_TOKEN=<personal access token with project read/write scope>
GITHUB_PROJECT_OWNER=<your GitHub username or org>
GITHUB_PROJECT_NUMBER=<the numeric ID of your GitHub Projects board>
The GITHUB_PROJECT_NUMBER is the number in the board URL:
https://github.com/users/<owner>/projects/<number>
Then start Machine Foundry:
python -m machinefoundryMachine Foundry will shape the board on first run and begin watching for items in the Ready column.
Any issue on the board can be moved to Ready manually. Machine Foundry will:
- Verify the issue has no unresolved blockers or incomplete sub-tasks.
- Dispatch an AI agent to work on it.
- Move the card to In Progress and record the agent session URL.
- Move the card to In Review when the agent is done.