Use AgentQL Skill in Claude Code
Claude Code is a good fit when you want the agent to inspect a codebase, answer integration questions, and validate hosted AgentQL calls in the same session.
Setup
Install the published skill
Install command
npx skills add AgentQL/skills -g -yExport one API key
Create the key on agentql.tomo.inc, then expose it to the Claude Code environment:
Export API key
export AQL_WORKSPACE_API_KEY="<workspace_api_key>"Keep endpoint truth externalized
Use
agentql.tomo.inc/nodesas the current endpoint source of truth.
Recommended tasks
- validate whether a feature should use Market HTTP, Market WebSocket, MCP, or EVM RPC
- fetch one live price, market profile, or short OHLCV window during feature design
- run one raw EVM RPC check before wiring an SDK or transport client
- keep prompt-level research separate from production application code
Prompt patterns
Use the installed AgentQL skill to decide whether this feature should call Market HTTP, MCP Service, or EVM RPC HTTP. Return one recommended surface and one example request shape.Use the installed AgentQL skill to read the latest price for WETH on ethereum, then summarize the response fields that matter for application code.Routing rule
- Product or market reads: Market HTTP
- Live market delivery: Market WebSocket
- Tool-first workflow: MCP Service
- Raw chain compatibility: EVM RPC HTTP or RPC WebSocket
Avoid
- Do not leave production hostnames hardcoded in prompts or code comments.
- Do not use raw RPC when a normalized market surface already answers the task.
- Do not treat the skill as the long-term runtime integration path for app code that should use the SDK.
Last updated on