Skip to content

MCP — Use BuildWorkPro from Claude and ChatGPT

The Model Context Protocol (MCP) is Anthropic’s open standard for connecting large language models to external tools and data. An MCP-aware client — Claude Desktop, ChatGPT, Cursor, your own agent — speaks one protocol, and any MCP server (BuildWorkPro included) can plug in. The model gets structured access to your records, with scoped permissions, instead of you copy-pasting screenshots into a chat.

The BuildWorkPro MCP server exposes the same domain you already use through the REST API, packaged for LLMs:

  • 24 tools across 7 entity domains plus identity. Read your contacts, leads, bids, projects, pay applications, change orders, and documents. Create new records and drive workflow steps (send a bid, accept a bid, approve a pay app, approve a change order).
  • 2 resources (bid:// and project:// URI schemes) that let the model fetch a single record by id with one round trip and pin it into context.
  • 3 prompts the model can call by name to compose useful workflows: draft_bid_followup, summarize_open_projects, and weekly_status_report.

Tools are scoped. Listing contacts requires contacts:read; creating a bid requires bids:write; sending a bid requires bids:send. The full inventory is in the Tools catalog.

The server uses OAuth 2.1 with Dynamic Client Registration (RFC 7591) — the same flow Zapier uses. You don’t paste an API key into your AI client. Instead:

  1. The MCP client (Claude, ChatGPT) registers itself with BuildWorkPro automatically the first time you connect.
  2. You’re redirected to the BuildWorkPro authorize page in your browser.
  3. You see a consent screen listing the exact scopes the AI is requesting and choose what to grant.
  4. The client receives a refresh token and exchanges it for short-lived access tokens.

Your AI assistant gets only the scopes you grant. Revoke access at any time from Settings -> Developer -> Connected apps.

  • Streamable HTTP (recommended). The hosted endpoint at https://app.buildworkpro.com/api/mcp works with any modern MCP client. No local process to run, no Node.js to install. This is the path documented in the install guides below.
  • stdio (local development only). A local Node.js process speaks MCP over stdin/stdout. Useful when you’re hacking on the server itself or running on a machine without internet access. We don’t recommend this for everyday use.