MCP Server

The vCon MCP Server — an open-source Model Context Protocol server that gives AI assistants standardized access to vCon conversation data.

The vCon MCP Server is the canonical way for AI assistants — Claude, Cursor, custom agents — to read, write, and reason about vCon conversation data. It speaks the Model Context Protocol over stdio or Streamable HTTP, exposes 37 tools grouped into seven categories, and is backed by a Supabase Postgres deployment with optional Redis caching and pgvector-based semantic search.

The code lives at vcon-dev/vcon-mcp. It's MIT-licensed, TypeScript, and ships as both an npm package (vcon-mcp) and a Docker image (public.ecr.aws/r4g1k2s3/vcon-dev/vcon-mcp:main).

When to use it

• You want Claude Desktop (or any MCP client) to query a vCon database in natural language.

• You're building an LLM agent that needs durable, structured access to conversation history.

• You have vCons in a conserver-managed Supabase backend and want a query layer that's safe to expose to language models.

• You need the May 2026 contract / discovery surface — stable response envelopes, cursor pagination, byte-budget enforcement — for production LLM workflows.

Documentation layout

Get started:

• What is the vCon MCP Server? — overview, requirements, getting started

• MCP and AI — how MCP and AI assistants connect

• Business Cases — when to reach for this

Use it:

• What the vCon MCP Server Can Do — narrative overview of capabilities

• Tool Reference — every tool, grouped by purpose

• Contract Tools — the May 2026 LLM-facing surface (vcon_fetch, vcon_search, etc.)

Run it:

• Transport and Deployment — auth, transport modes, environment variables, Docker, npm, Claude Desktop config

• Field-Name Migration — how the server handles legacy appended / must_support field names

Understand it:

• How the vCon MCP Server is Built — architecture, schema, request flow

• MCP and Conserver Together — how the MCP server and conserver compose

External:

• GitHub: vcon-dev/vcon-mcp — source

• mcp.conserver.io — auto-generated technical reference