Skip to main content
Coding agents need a market-data API that exposes route families, schemas, auth, freshness, and examples clearly enough to generate safe requests without guessing. 0xArchive does, across one OpenAPI contract and four execution surfaces. The route map comes from OpenAPI and docs; execution comes from the CLI for shell jobs, SDKs for application code, the MCP Server for typed tool calls, and Skills for reusable agent workflows. One key returns Hyperliquid and Lighter market data, live and historical: every order, trade, and fill.

​
What One Key Gives An Agent

Give agents the smallest exact surface that matches the job. The agent host and the tool surface are different decisions.
Agent environment0xArchive surfaceWhy
Web LLMDocs Markdown, .md pages, llms.txt, OpenAPI excerptsSafe retrieval and route explanation
Coding agent: Claude Code, Codex, Cursor, Windsurf, Copilot, Gemini CLI, Devin, OpenCodeOpenAPI and docs first; CLI, SDK, MCP Server, or Skill as neededCode generation plus local execution patterns
OpenClaw harnessSkill for simple market-data prompts; CLI, MCP Server, or ACP when the setup is configured for those surfacesInstall skills, run external agents, or bridge tools per job
Automation scriptREST, SDK, CLI JSON outputRepeatable jobs with environment-managed keys
MCP-capable hostMCP ServerStructured tool calls without prompt-pasted secrets
Generated clientPinned openapi.jsonRoute and schema contract

​
Why Teams Choose 0xArchive

Market-data agents need more than a prose description. They need route families, parameter shapes, error handling, request IDs, freshness rules, and safe key handling, and 0xArchive exposes those primitives directly: /v1/hyperliquid/*, /v1/hyperliquid/spot/*, /v1/hyperliquid/hip3/*, /v1/hyperliquid/hip4/*, /v1/lighter/*, /v1/data-quality/*, the X-API-Key header, a generated OpenAPI reference, and agent-specific instructions in Skill discovery indexes and assistant context pages. An agent can read the OpenAPI route shape, send a bounded request, preserve meta.request_id, and check freshness before trusting the result, all from public docs.

​
Agent Selection Checklist

Before an agent executes, require a checklist with the job type, venue family, symbol format, route or channel, auth source, endpoint-specific response contract, freshness check, retry policy, request limit, and where output will be stored. If the agent can fill that checklist from public docs and OpenAPI, it can execute; if not, keep it in explanation mode.

​
Agent Prompt Pattern

​
Next Step

Open Choose an interface for the surface matrix, then use AI and coding agents and CLI for setup paths. Compare plans.
Last modified on June 28, 2026