Skip to main content
Version: Current

Agentic Trading

Agentic Trading lets AI clients inspect Nunchi markets and, when explicitly authorized, request trading actions through scoped permissions.

Start with read-only access, verify the agent's behavior on low-risk workflows, and only enable write tools after you understand the permissions you are granting.

Setup​

Use the hosted MCP endpoint to connect Cursor, Claude, Codex, ChatGPT, Grok, or another MCP-capable client.

See Agentic Trading MCP for:

  • MCP URL and client setup steps.
  • Permission tiers.
  • Supported clients, including Grok.
  • Risk notes before enabling write access.

What Agents Can Do​

Depending on the permissions you grant, an agent can:

  • Inspect portfolio, market, strategy, and status data.
  • Summarize open positions, recent fills, and risk concentration.
  • Run testnet trading workflows.
  • Request live trading actions only after explicit mainnet and write-tool consent.

The hosted runner does not handle your private keys.

Disclosures​

Before enabling live trading, review the Agentic Trading Beta Disclosures. You remain responsible for every order your AI agent places.