vendus-mcp 0.3.0

Model Context Protocol server for the Vendus AI sales platform

vendus-mcp

Model Context Protocol server for the Vendus AI sales platform. It lets any MCP-aware client (Claude Code, Claude Desktop, Cursor, custom agents) read leads, conversations, agents, metrics and knowledge base documents from a Vendus tenant via the public REST API.

V1 is read-only and scoped per API key.

Quick start

# 1. Create an API key in your Vendus dashboard (Settings > API Keys)
# 2. Run the MCP server with uvx — no install required
VENDUS_API_KEY=vnd_live_xxx \
  VENDUS_BASE_URL=https://your-vendus-instance.fly.dev \
  uvx vendus-mcp

The server speaks MCP over stdio, so any compliant client can mount it.

Configuration

Env var	Required	Default	Description
VENDUS_API_KEY	yes	—	API key issued from Settings > API Keys
VENDUS_BASE_URL	no	https://semory-dashboard-backend.fly.dev	Base URL of your Vendus backend

Keys are tenant-scoped and bound to a set of read scopes. Calls fail closed when a scope is missing — there is no implicit elevation.

Use it from Claude Code

Add the server to your project's .mcp.json (or to the global Claude Code config):

{
  "mcpServers": {
    "vendus": {
      "command": "uvx",
      "args": ["vendus-mcp"],
      "env": {
        "VENDUS_API_KEY": "vnd_live_xxx",
        "VENDUS_BASE_URL": "https://your-vendus-instance.fly.dev"
      }
    }
  }
}

Then run claude mcp list and you should see vendus reporting ✓ Connected. The tools become available inside any Claude Code session in that project.

Use it from Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) and add the same mcpServers.vendus block. Restart Claude Desktop.

Use it from Cursor / Continue / other MCP clients

Any client that can launch a stdio MCP server works. Point it at uvx vendus-mcp with the env vars above.

Host it for Claude.ai Custom Connectors and ChatGPT remote MCP

Claude.ai and ChatGPT both accept a hosted MCP endpoint instead of a stdio binary. The same package ships a streamable-HTTP variant:

VENDUS_BASE_URL=https://your-vendus-instance.fly.dev \
  uvx vendus-mcp-http --host 0.0.0.0 --port 8080 --mount-path /mcp

Authentication is per-request: the client (Claude / ChatGPT) forwards the user's token in Authorization: Bearer <vnd_live_...> on every MCP request, so a single deployment safely serves many tenants. VENDUS_API_KEY is not read in HTTP mode — each caller brings their own.

Register the endpoint:

• Claude.ai: Settings → Connectors → Add custom connector → URL https://your-host/mcp, auth Bearer with the user's Vendus API key.
• ChatGPT: Settings → Apps & Connectors → Add remote MCP server → same URL + Bearer.

For the path to the official connector directories (OAuth 2.1 + DCR + business registration), see docs/superpowers/specs/2026-04-17-mcp-connector-directory.md.

Tools

Tool	Description
list_leads	Filter by status, qualified, search, date range
get_lead	Lead profile by id
get_conversations	Chronological messages for a lead
list_agents	All AI agents on the tenant
get_agent	Agent profile by id
get_metrics_summary	Funnel counters since a date
get_metrics_funnel	Conversion percentages per stage
list_knowledge_documents	Knowledge base documents, paginated
get_knowledge_document	Document metadata by id

All tools return JSON strings shaped like {"data": ...} or {"error": {"code": "...", "message": "...", "status": 4xx}} on failure.

Local install (for hacking)

git clone https://github.com/Agentibus-ai/vendus
cd vendus/mcp-server
pip install -e ".[dev]"
pytest

Roadmap

• Write tools (leads:write, conversations:write) once the REST API exposes them.
• HTTP-streamable transport for hosted Anthropic / OpenAI connectors.
• Webhook subscription tools.

License

MIT — see LICENSE.