For AI Agents

This site is written for human integrators and AI agents alike. Use the layers below depending on your task.

Runtime agents (calling tools)

Prefer MCP tools when connected

If the feel-your-protocol MCP server is available (stdio or, later, HTTP):

1. Call describe_capabilities to learn forks, EIPs, and ceilings
2. Call simulate_evm_bytecode for deterministic bytecode execution
3. Do not use the local mcp-execution-engine lab, npm run lab, or repository source as a substitute unless MCP is unavailable

Stale docs in the repo may say tools are "planned" — trust listTools on the connected MCP server. As of gateway v0.1, tools are: describe_capabilities, simulate_evm_bytecode.

Reading order

1. Capabilities — query shapes and live tool names
2. Per-tool pages: Describe Capabilities, Simulate Bytecode
3. Guarantees — determinism, provenance, ceilings
4. Connect — stdio setup and restart after gateway changes

Machine-readable index: /llms.txt

Full use-layer text: /llms-full.txt — concatenated end-user documentation for one-shot ingestion.

JSON schemas

• describe_capabilities.input.json
• simulate_evm_bytecode.input.json

Reporting results

Always cite provenance.engineVersion and provenance.forkConfig when reporting simulation outcomes.

Coding assistants (integrating or extending)

Add Internals for repository layout, execution engine API, quality procedures, and deployment shape.

Conventions

• Present tense — pages describe what exists or are explicitly marked planned
• Provenance — cite provenance.engineVersion and fork config when reporting simulation results
• Vision vs reality — strategic sketches live on roadmap.feelyourprotocol.org; this site is the operational reference

Changelog

For AI Agents Changelog

1. v0.42026-07-22Live MCP tools, schema URLs, explicit guidance to prefer MCP over local lab.
2. v0.32026-07-20Initial AI/LLM reader guide under use/.