Coffee for AI Agents

    The customer data layer for your AI agents.

    Building GTM workflows with Claude Code, Codex or Cowork? Don't roll your own storage, dedupe and history. Plug into Coffee.

    Start freeBook a demo

    MCP on day one

    Every user with OAuth access to Coffee, including free-trial users, can connect Coffee to Claude, ChatGPT, Claude Code, Codex or any MCP-compatible agent. No setup ticket, no approval.

    What agents can do

    Read records (every deal in a date range, a deal's full history), create and update records, and run operational or analytics workflows, with lineage on every record.

    Remote control your CRM

    A sales leader asks Claude what changed in pipeline this week. A coding agent writes clean records into Coffee as part of a custom workflow.

    API for machine-to-machine

    For M2M use cases, Coffee provisions a token with you. MCP is self-serve; the API is set up together.

    Connect in one step.

    Sign in with OAuth from any MCP-compatible client.

    # Connection config
{
  "mcpServers": {
    "coffee": {
      "url": "COFFEE_MCP_URL"
    }
  }
}

Try: "Show me every deal that moved backward a stage in Q3 and why"

    Answers for builders and their agents

    In a system that keeps it clean and keeps its history. Coffee stores contacts, companies, deals and activity in a data lakehouse with full lineage, so agents get authoritative data without you building storage. Connect through Coffee's MCP server.

    Coffee merges duplicates at write time, including the same person with a new email address but matching details, and keeps contact-to-account associations current. Every merge is logged and reversible.

    Coffee records every change to every field, with its source. Agents can request a record's full history through MCP.

    Add Coffee as an MCP server in Claude Code and sign in with OAuth. Every Coffee user, including free trials, has access on day one.