Pegboard
Docs menu

Teach your agents

Install the built-in skill so your agents use Pegboard instead of editing config files.

Pegboard ships with a skill that teaches agents to use the CLI. With it, “add the Linear MCP to all my agents” takes a couple of deterministic commands instead of an agent reading and rewriting several formats.

Install it

pegboard skill install

This installs the skill in the global scope of every detected agent. Pass -y to skip the confirmation. To remove it:

pegboard skill uninstall

You can also declare it in a stack, like any other skill:

skills:
  - pegboard

What it tells agents to do

  1. Never edit config files by hand. That covers .mcp.json, .codex/config.toml, the skills directories, and the rest.
  2. Plan first. Run the command with --dry-run --json, summarize the changes for you, and only then repeat it with -y --json.
  3. Ask about conflicts. Explain the reason and ask you before using --force.
  4. Never write secrets in plain text. Use references, and tell you if an agent can’t support them.

Which commands they use

To copy your setup from any agent to others, the agent uses replicate:

pegboard replicate claude-code codex opencode --dry-run --json
pegboard replicate claude-code codex opencode -y --json

If the project has a pegboard.yaml, the agent edits the stack and applies it:

pegboard add mcp notion=https://mcp.notion.com/mcp --agent claude-code
pegboard plan --json
pegboard apply -y --json

If it doesn’t, the agent uses ad-hoc mode:

pegboard install mcp notion=https://mcp.notion.com/mcp --dry-run --json
pegboard install mcp notion=https://mcp.notion.com/mcp --json -y