| name | doey-mcp-example |
| description | Example skill demonstrating MCP server integration. Not for actual use. |
| mcp_servers | [{"name":"filesystem","command":"npx","args":"-y @modelcontextprotocol/server-filesystem /tmp","env":{}},{"name":"github","command":"github-mcp-server","args":"stdio","env":{"GITHUB_PERSONAL_ACCESS_TOKEN":"${GITHUB_TOKEN}"}}] |
MCP Example Skill
This is a reference skill demonstrating the mcp_servers YAML frontmatter field.
How MCP Integration Works
Skills can declare MCP server dependencies in their YAML frontmatter using the mcp_servers field.
When a worker is launched to execute a skill with MCP dependencies, Doey:
- Parses the
mcp_servers block from the skill's SKILL.md frontmatter
- Generates a per-pane
.mcp.json config file in $RUNTIME_DIR/mcp/
- Passes
--mcp-config <path> when launching the Claude instance
- Cleans up MCP configs and processes on team despawn
Frontmatter Schema
mcp_servers:
- name: <server-name>
command: <executable>
args: <arg1> <arg2> ...
env:
KEY: value
KEY2: ${ENV_VAR}
Limitations
- MCP servers start at Claude launch — they cannot be added/removed at runtime
- Each team gets independent MCP server instances (no cross-team sharing)
- MCP tools appear as
mcp__<server-name>__<tool-name> in tool listings