PUBLIC KNOWLEDGE ENDPOINT / READ ONLY

One endpoint. Three AI coding worlds.

Add a source-labeled CLI and editorial reference to Claude Code, Codex, or Grok Build. It keeps each product separate, returns bounded evidence, and never runs the commands it finds.

Hosted endpointhttps://mcp.claude-world.com/mcpHTTP
01 / CONNECT

Connect your client

Run the add command for your client, then the adjacent registration or connectivity check.

Claude Code

Add
claude mcp add --transport http --scope user worlds-knowledge https://mcp.claude-world.com/mcp
Check
claude mcp get worlds-knowledge

Codex

Add
codex mcp add worlds-knowledge --url https://mcp.claude-world.com/mcp
Check
codex mcp get worlds-knowledge --json

Grok Build

Add
grok mcp add --transport http --scope user worlds-knowledge https://mcp.claude-world.com/mcp
Check
grok mcp doctor worlds-knowledge --json
02 / TEST

Make the first query count

Paste this after connecting. It exercises cross-product separation, release metadata, and returned citations.

Use the AI Worlds Knowledge MCP to compare what -p means in Claude Code, Codex, and Grok Build. Keep the products separate, include each captured release, and cite the returned official sources.
03 / BOUNDARY

Read-only by design

  • Your query text and selected filters are sent to the hosted endpoint. Do not include secrets, credentials, or private code.
  • The MCP exposes no local-file, local-setting-value, write, command-execution, or arbitrary-URL-fetch tool.
  • Returned commands are reference data. Your agent and approval policy still decide whether anything runs.
04 / SOURCES

A visible source policy

  • Released runtime evidence is preferred, then official documentation; unreleased source evidence stays labeled as preview.
  • Results preserve product, captured release, source status, review date, canonical link, and official sources.
  • Runtime, documentation, and source disagreements remain visible as drift. Recheck consequential guidance against the returned primary source.