Skip to main content
Install the 0xArchive Agent skill in Claude Code, Codex, OpenClaw, or another agent that loads SKILL.md workflows. With it, the agent picks the right 0xArchive route, keeps venue families apart, runs bounded curl requests, reads the JSON with jq, and keeps each request ID. Source: 0xArchiveIO/0xarchive-skill.

​
Requirements

​
Install

OpenClaw installs the package and makes it available to its agents.

​
Set the key

Put the API key in the shell or project environment the agent runs commands in:
The skill’s HTTP calls send it as the X-API-Key header. Do not paste the key into a prompt, and do not ask the agent to print it.

​
Run one prompt

Ask for one market, one data type, and a small result, using the call syntax from your agent’s install steps:

Use the 0xArchive skill to get the current BTC funding rate on Hyperliquid. Show the route you called and the request ID.

Good prompts name the venue family, symbol, data family, and output goal. Two more first prompts:

Use the 0xArchive skill to check whether Hyperliquid Spot HYPE-USDC has fresh trade data, then show one bounded request and the fields I should log.

Use the 0xArchive skill for HIP-3 km:US500. Choose the route family, run a small freshness or trades check, and keep the request ID.

​
Confirm success

A working setup answers with current or recent 0xArchive data for the market you asked about, and can show the route it called and the response’s meta.request_id. A missing or wrong key comes back as a direct authentication error: HTTP 401 with error_code unauthorized and a top-level request_id. Set OXARCHIVE_API_KEY in the environment the agent uses and ask again.

​
What the skill does

1

Choose the venue family

Separate Hyperliquid core, Hyperliquid Spot, HIP-3, HIP-4, and Lighter before choosing routes.
2

Run bounded requests

Use shell-backed curl patterns with explicit symbols, windows, limits, and auth headers.
3

Parse and gate output

Use jq to inspect success, data shape, freshness, and the request ID before expanding the job.
For a prompt such as “Check freshness for Lighter BTC, then pull the first 50 trades if the freshness check is acceptable,” the skill can choose GET /v1/lighter/freshness/BTC, then GET /v1/lighter/trades/BTC, use the key safely, parse each response contract, and preserve the request ID. Resolve any additional parameters in OpenAPI before calling.

​
When another tool fits better

The skill suits an agent that already has a shell and needs repeatable market-data steps rather than a generated client. Use the MCP server when your client calls tools, the CLI when the work should become a command for scripts, CI, or cron, and an SDK or REST when the code belongs inside an application. OpenAPI has the route and schema contract for each.

​
Machine discovery

Agents and tools that find skills automatically read these indexes on the docs site:
  • https://docs.0xarchive.io/.well-known/skills/index.json: the skill index, with a description and file list for each skill.
  • https://docs.0xarchive.io/.well-known/agent-skills/index.json: the agent-skill index, with each skill’s URL, description, and SHA-256 digest.
  • The skill files that index points to: https://docs.0xarchive.io/.well-known/agent-skills/0xarchive-market-data/skill.md, https://docs.0xarchive.io/.well-known/agent-skills/0xarchive-openapi/skill.md, and https://docs.0xarchive.io/.well-known/agent-skills/0xarchive-websocket/skill.md.
Do not treat /skill.md as the public Markdown Skill package: that address can serve generated discovery data instead. Use the indexes above.

​
Next step

Give the agent jobs it can finish with the job checklist.
Last modified on October 6, 2026