Skip to main content

​
How it works

  1. Your agent calls a Nebula tool over MCP (local nebulamcp-stdio or remote POST /mcp).
  2. The Hub authenticates the nbl_live_… token (or OAuth) and loads that agent’s wallet and policy for its network.
  3. Policy runs (caps, allow/deny, pause). Some actions return confirmation_required until you approve in the dashboard.
  4. The Hub signs with Privy custody and submits to Stellar. The agent never sees a secret key.
Dashboard-only actions (policy writes, Blend deposit/withdraw, liquidity threshold) require your Hub login. Agents cannot call them via MCP. See Custody, Spending policy, and Confirmations.