pegboard.yaml: Pegboard docs

A stack lives at ./pegboard.yaml for a project and ~/.config/pegboard/pegboard.yaml for your user. It’s validated against a JSON Schema, which also gives you autocompletion in editors that support it.

version: 1
agents: [claude-code, codex, opencode, hermes]
inherit: global

mcps:
  github:
    command: npx
    args: ["-y", "@modelcontextprotocol/server-github"]
    env:
      GITHUB_TOKEN: "${env:GITHUB_TOKEN}"
  notion:
    url: https://mcp.notion.com/mcp
    agents: [claude-code]

skills:
  - mattpocock/skills@tdd
  - pegboard
  - source: ./skills/review
    agents: [codex, opencode]

context:
  agents_md: true

remove:
  mcps: [notion]

policy:
  allow_mcps: [github, linear]

Keys

Key	What it does
`version`	Format version. Use `1`
`agents`	Which agents to configure. Defaults to the ones detected. See agent IDs
`inherit`	Project stacks only. `global` or `none`
`mcps`	MCP servers, by name
`skills`	Skill sources. To limit one to some agents, write `source: …` and `agents: […]`
`context.agents_md`	Project scope only. Guarantees `AGENTS.md` and the `CLAUDE.md` bridge
`remove`	Removes an entry that’s declared or inherited, applied last
`policy.allow_mcps`	Fails `plan` and `apply` if any other MCP is present
`include`	Planned. Currently ignored with a warning

MCP entries

Key	What it does
`command`, `args`, `env`	A stdio server
`url`	An HTTP server
`agents`	Limits this MCP to certain agents

`context.agents_md`

This creates AGENTS.md if it doesn’t exist. From then on the file is yours, and Pegboard never deletes it. It also adds the line @AGENTS.md to CLAUDE.md so Claude Code reads it. If you later remove agents_md, Pegboard removes that line only if it added it.