Agent API

Every StockKit endpoint is designed to be agent-friendly: plain JSON over HTTPS, open CORS, no authentication during the beta, and stable machine- readable shapes.

Discovery

GET https://api.stockkit.dev/ returns a machine-readable index of all endpoints. From there an agent can:

• Discover: GET /v1/assets enumerates every tradable tokenized asset with onchain addresses
• Price: GET /v1/prices/:symbol returns live quotes with volume, daily range, and halt status
• Research: GET /v1/research/:symbol returns the SEC-registered company behind a token with as-filed fundamentals
• Quote: GET /v1/quote returns executable DEX quotes
• Prepare trades: POST /v1/trade/build returns unsigned transactions for the agent’s own wallet to sign
• Manage capital: GET /v1/portfolio/:address values any wallet’s positions in USD
• Chart: GET /v1/history/:symbol returns OHLC candles of the onchain DEX price; GET /v1/stats/:symbol returns supply, pool liquidity, and the DEX premium vs the equity quote
• Watch: POST /v1/alerts registers a one-shot price alert delivered to the agent’s webhook

Paying per call

Agents that would rather pay in-protocol than use the open beta endpoints can call the same capabilities under /v1/x402/*, priced per call in USDG on Robinhood Chain or USDC on Base. See the x402 API.

Non-custodial by design

Agents hold their own keys. StockKit only prepares calldata; the agent signs and submits with its own wallet infrastructure and remains fully in control of funds.

Errors

All errors share one shape, so agents can branch reliably:

{ "error": { "code": "no_liquidity", "message": "No executable route for XYZ" } }

Codes: bad_request, not_found, no_liquidity, upstream_error, internal_error.

MCP server

StockKit is also a hosted MCP server at https://api.stockkit.dev/mcp, so Cursor, Claude Code, and any MCP client can use these capabilities as native tools. It ships as a one-click Cursor install (stockkit-labs/cursor) and a Codex plugin (stockkit-labs/openai). See Use with Cursor and Claude Code.