# Model Context Protocol (MCP)
Decapod is designed to be a "Protocol First" governance layer. While it provides a rich CLI, its core value is in standardizing the context and control signals exchanged between repositories and AI models. This aligns perfectly with the **Model Context Protocol (MCP)**.
## Decapod as an MCP Provider
Decapod integrates with MCP-compatible environments (like Claude Desktop or specialized IDE extensions) by acting as an MCP Server. This allows agents to access Decapod's governance kernel directly through standard protocol messages.
### Exposed MCP Resources
- **Constitution:** Browsable nodes of the technology and methodology graph.
- **Tasks:** The current todo list and ownership state.
- **Specs:** The living specifications (`INTENT.md`, `ARCHITECTURE.md`, etc.) for the project.
- **Context Capsules:** Deterministic snapshots of repo state used for inference.
### Exposed MCP Tools
Decapod exposes its core CLI capabilities as MCP Tools, allowing agents to:
- `todo_add` / `todo_claim`
- `workspace_ensure`
- `constitution_search`
- `aptitude_resolve_skills`
- `validate`
## Benefits of MCP Integration
1. **Lower Friction:** Agents don't need to "learn" CLI flags if they can call structured tools via MCP.
2. **Rich Context:** MCP allows for richer metadata (like URIs and structured resources) to be passed alongside documentation.
3. **Discovery:** MCP-native agents can automatically discover Decapod's capabilities as soon as they connect to the repository.
## Handshaking and Identity
Decapod uses its `handshake` command to generate a deterministic identity artifact. In an MCP context, this ensures that the agent provider and the local governance kernel have a shared, cryptographically verifiable understanding of which agent is operating on which task.
## Future Direction
We are actively expanding Decapod's MCP surface to support native resource templates and more granular tool definitions, making Decapod the standard "Governance MCP" for any repository it manages.