Protocols
Vectora implements two communication protocols: MCP (Model Context Protocol) for integration with IDEs, and ACP (Agent Communication Protocol) for communication between sub-agents.
Supported Protocols
| Protocol | Use Case | Status | Docs |
|---|---|---|---|
| MCP | Connection with Claude Code, Cursor | Stable | → MCP |
| ACP | Communication between sub-agents | Beta | → ACP |
MCP (Model Context Protocol)
The standard protocol for modern IDEs.
MCP is an open protocol developed by Anthropic that allows LLMs to call tools on a computer in a structured way. Vectora offers 12 tools via MCP.
Advantages:
- Native in Claude Code and Cursor (zero extra config)
- Dynamic tool discovery
- Schema validation (ZOD)
- Automatic result caching
- Latency <10ms (local IPC)
ACP (Agent Communication Protocol)
For communication between Vectora and custom agents.
ACP allows multiple agents to work together, sharing context and state. Ideal for distributed architectures where Vectora is a sub-agent of a larger system.
Status: Beta - Available for early adopters
Which Protocol to Use?
- Using Claude Code / Cursor / Zed? → MCP
- Integrating with custom agent in Python/Node/Go? → MCP or REST API (beta)
- Agent-to-agent communication? → ACP (beta)
See MCP for full implementation details.
Next: MCP Specification
External Linking
| Concept | Resource | Link |
|---|---|---|
| MCP | Model Context Protocol Specification | modelcontextprotocol.io/specification |
| MCP Go SDK | Go SDK for MCP (mark3labs) | github.com/mark3labs/mcp-go |
| Anthropic Claude | Claude Documentation | docs.anthropic.com/ |
| Zod | TypeScript-first schema validation | zod.dev/ |
Part of the Vectora ecosystem · Open Source (MIT) · Contributors