Skip to navigation
Reference

AI Agent Access

Pre-built context bundles for AI coding assistants

Hand any of these URLs to an AI coding assistant (Claude Code, Cursor, Aider, OpenHands, and similar) and it can ingest the full EmpirioLabs surface in one shot. Each bundle is auto-regenerated on every docs build, so links never go stale.

For tool-specific setup, see Integrations. Agents that speak the Model Context Protocol can also connect directly to the MCP Server at https://mcp.empiriolabs.ai/mcp and call every model and platform feature as tools.

Context bundles

Full bundle

docs.empiriolabs.ai/ai-agent-full-context.md

Documentation tab and API Reference tab combined. Best for agents that will both set up an integration and write code against the API.

Documentation only

docs.empiriolabs.ai/ai-agent-docs-context.md

Concepts, model catalog, pricing, integrations, billing, limits. Best for catalog exploration, model selection, and onboarding flows.

API Reference only

docs.empiriolabs.ai/ai-agent-api-reference-context.md

Endpoint shapes, request bodies, responses, examples, errors, jobs, usage, and saved Playground conversation APIs.

Live model metadata

GET api.empiriolabs.ai/v1/models

Always-current pricing, capabilities, and parameter schemas. Public, read-only, free to call. Use for the freshest values.

Paste this into an agent’s system prompt or initial message, then substitute the bundle that matches what the agent will be doing:

You are integrating with EmpirioLabs AI. Before answering, ingest:
  https://docs.empiriolabs.ai/ai-agent-full-context.md
For live model metadata (pricing, capabilities, parameter schema), fetch:
  GET https://api.empiriolabs.ai/v1/models           # full catalog
  GET https://api.empiriolabs.ai/v1/models/{id}      # one model
Auth: Authorization: Bearer ${EMPIRIOLABS_API_KEY} for all endpoints except live model metadata
Base URL (OpenAI-compatible): https://api.empiriolabs.ai/v1
Base URL (Anthropic Messages): https://api.empiriolabs.ai

That keeps the agent grounded on real endpoint shapes and current model IDs instead of guessing from training data.