AI Hub Roadmap

Review the recommended staged rollout for public data, account read, preview, controlled trading, and skills.
View as Markdown

AI Hub should be released in stages so users receive value early while sensitive trading capabilities remain protected.

Phase 1: Public Market AI Access

Goal: let any supported AI client read official 6MM market data.

Deliverables:

  • Shared SDK / API client.
  • CLI public market commands.
  • Local MCP public market tools.
  • Hosted MCP public market tools.
  • Resources for markets, fees, risk rules, and system status.
  • Client installation guides.

Not included:

  • Account queries.
  • Real trading.
  • API key hosting.

Phase 2: Account Read

Goal: let users safely inspect their own account context.

Deliverables:

  • Read-only API key scope.
  • Account summary.
  • Balances.
  • Positions.
  • Open orders.
  • Fills.
  • Funding payments.
  • Portfolio risk prompt.

Preferred deployment:

  • Local MCP.
  • CLI.
  • Remote MCP self-hosted.

Phase 3: Trading Preview

Goal: let AI help generate orders without executing them.

Deliverables:

  • preview_order.
  • Risk checks.
  • Margin impact.
  • Fee estimate.
  • Liquidation impact.
  • Execution planning prompt.

Phase 4: Controlled Trading

Goal: support real trading only with strong guardrails.

Deliverables:

  • create_order.
  • cancel_order.
  • amend_order.
  • cancel_all_orders.
  • order_status.
  • idempotency.
  • max notional limits.
  • explicit confirmation.
  • audit logs.

Phase 5: Agentic Skills

Goal: productize common market analysis and risk workflows.

Deliverables:

  • Market analyst skill.
  • Portfolio risk skill.
  • Funding research skill.
  • Liquidation protection skill.
  • Execution planner skill.
  • Example task library.