Skip to main content
Bravado is built to be consumed by AI-native tooling. This page shows how to point a coding assistant, an autonomous agent, or an MCP-aware client at the Bravado docs and API.

​
Machine-readable docs

Every page on this site is available as raw Markdown:
  • Append .md to any docs URL to get the raw source. For example, https://bravado.mintlify.app/quickstart.md.
  • https://bravado.mintlify.app/llms.txt lists every page for LLM crawlers.
  • https://bravado.mintlify.app/llms-full.txt concatenates the entire docs site in one file, suitable for pasting into a long-context model.
Point Claude, Cursor, or any other assistant at either URL when you want it to have full context on Bravado without scraping.

​
Copy any page as Markdown or open in an assistant

Every page has a Copy button in the contextual menu. Options include:
  • Copy page as Markdown for pasting into a prompt.
  • View as Markdown to see the raw source in a new tab.
  • Open in ChatGPT to load the page into ChatGPT.
  • Open in Claude to load the page into Claude.

​
Pointing a coding agent at Bravado

For Cursor, Windsurf, or Claude Code, add the docs to the workspace context:
The agent will then have the entire API surface, all concepts, and all guides in its context window when generating Bravado code. For ChatGPT or Claude web, use the “Open in ChatGPT” / “Open in Claude” button on any specific page you need the model to reason about.

​
MCP server

Bravado will expose a hosted Model Context Protocol (MCP) server so agent frameworks that speak MCP (Claude Desktop, some IDE extensions, custom agent runtimes) can call Bravado tools directly without a REST wrapper. Expected tools include:
  • place_order, cancel_order, get_positions, get_balances on the Trade surface.
  • subscribe_copytrade, unsubscribe_copytrade on the Copytrade surface.
  • get_leaderboard, get_trader_pnl, get_trader_positions on the Analytics surface.

​
Safety patterns for autonomous agents

When wiring an agent to Bravado’s trading endpoints:
  1. Cap notional per tool call. Enforce the cap in your tool wrapper, not the prompt.
  2. Require idempotency at the tool layer. Generate the UUID inside the tool. See Idempotency.
  3. Sandbox first. Test agents against a follower wallet with a small balance.
  4. Log every call. Every mutating call includes an idempotency key. Log per-agent, per-session for review.
See the AI agents use case for a worked example.