Skip to Content
AgentQL SkillIDE AgentsUse in Cursor

Use AgentQL Skill in Cursor

Cursor works best when you separate agent-side exploration from production application integration.

  • Use the AgentQL skill when Cursor is answering questions, testing a hosted call, or preparing example requests.
  • Use the Typescript-SDK when your application code should call AgentQL directly.
  • Use the transport docs when you need exact HTTP, WebSocket, MCP, or RPC request shapes.

Minimum setup

Make the API key available

Create the key on agentql.tomo.inc, then expose it to the Cursor agent or integrated terminal:

Export API key
export AQL_WORKSPACE_API_KEY="<workspace_api_key>"

Install the skill when the host supports it

Install command
npx skills add AgentQL/skills -g -y

If the current Cursor workflow does not expose the same install path, keep the same prompt contract and call the documented hosted surfaces directly.

Keep route truth outside the prompt

Load current public endpoints from agentql.tomo.inc/nodes.

When Cursor should use Skill

  • compare two assets or markets quickly
  • fetch one price snapshot or short OHLCV window
  • validate an EVM RPC method before wiring app code
  • draft example calls for docs, tests, or demos

When Cursor should use SDK or raw docs instead

  • the feature being built must call AgentQL at runtime
  • the repo needs typed application code instead of an agent answer
  • the flow depends on a live websocket client or explicit request headers

Prompt patterns

Use AgentQL to read a 15m OHLCV window for this ethereum token and return the last 3 candles as JSON.
Use AgentQL to validate whether this feature should call Market HTTP, EVM RPC HTTP, or Market WebSocket. Answer with one recommended surface and one example request shape.

Decision rule

Last updated on
Use in CodexUse in Windsurf