Learn

The desk, in writing

Gambit is an AI research desk for Hyperliquid. It initiates with a ranked opportunity feed and a thesis card — evidence, counterpoint, invalidation, size — and waits for you to confirm. Agents connect over MCP at api.gambit.trading/mcp: sign in with OAuth, or use a scoped API key for headless scripts.

Updated 6 October 2026 · Gambit

  1. AI tools for Hyperliquid traders

    Hyperliquid traders currently find dashboards, funding screens, community MCP wrappers, and a few AI desks. Gambit is the research desk that initiates: it ranks opportunities, writes a thesis with a counterpoint and an invalidation, and waits for a human confirm. It does not trade unattended and it does not replace a charting terminal.

  2. Gambit FAQ — custody, confirms, and fees

    Gambit is not a custodian and not a broker. The desk prepares a ticket; you confirm. In the app that confirm is the Trade surface. An MCP connection or API key granted api:trade can propose and confirm after the client shows you the preview — treat it as a live trading credential. Without api:trade it cannot place trades. There are no guaranteed returns.

  3. Connect an agent to Gambit (MCP)

    Gambit speaks MCP at https://api.gambit.trading/mcp over Streamable HTTP. Add that URL to Claude, ChatGPT, Claude Code, Codex, Cursor, or VS Code and sign in with OAuth; you choose the scopes on the consent page. Headless agents use a scoped API key instead. propose_trade and confirm_trade need api:trade and must show the preview to a human before confirm.

  4. What is a trade thesis card

    A trade thesis card is the desk’s unit of work. It states the instrument and direction, the evidence that initiated it, the counterpoint that would make it wrong, the invalidation, and a size that already respects your policy. It is a decision object, not a chat transcript and not a signal blast.

  5. What would prove this trade wrong

    An invalidation is the condition that retires a thesis. Gambit writes it on the card next to the evidence: a price, a failed hold, a catalyst that did not print. If that condition hits, the idea is done. The point is to decide in advance what would prove you wrong, not to invent a new story after the tape moves.