One profile. Every agent runtime.

Drop a more.md profile into Cursor via MCP. The agent can read, search, pay-per-read any profile from the editor.

// .cursor/mcp.json
{
  "mcpServers": {
    "more-md": {
      "command": "npx",
      "args": ["-y", "@more-md/mcp@latest"],
      "env": { "MORE_MD_USER": "<your-username>" }
    }
  }
}

Same MCP stdio server, wired into Anthropic's Claude Desktop config file.

// ~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "more-md": {
      "command": "npx",
      "args": ["-y", "@more-md/mcp@latest"],
      "env": { "MORE_MD_USER": "<your-username>" }
    }
  }
}

Codeium Windsurf speaks MCP too. Same stdio server, dropped into the Windsurf config.

// ~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "more-md": {
      "command": "npx",
      "args": ["-y", "@more-md/mcp@latest"]
    }
  }
}

OpenAI Codex CLI loads MCP from a TOML block. Ships with rate-limited read access out of the box.

# ~/.codex/config.toml
[mcp_servers.more-md]
command = "npx"
args = ["-y", "@more-md/mcp@latest"]

Install the more-md MCP server through the VS Code MCP bridge (agentic chat panel).

// .vscode/mcp.json
{
  "servers": {
    "more-md": {
      "command": "npx",
      "args": ["-y", "@more-md/mcp@latest"]
    }
  }
}

Antigravity reads the same MCP schema. Copy the config in, restart the runtime, done.

// ~/.antigravity/mcp.json
{
  "mcpServers": {
    "more-md": {
      "command": "npx",
      "args": ["-y", "@more-md/mcp@latest"]
    }
  }
}

First-class Zed MCP integration lands in W2 once the upstream context-server API stabilizes.

Typed adapter ships in Wave 2. Use the framework's generic HTTP tool today.

Continue's MCP bridge ships in W2; until then you can wire more.md via its generic HTTP tool.

Typed adapter ships in Wave 2. Use the framework's generic HTTP tool today.

Load a profile as a LangChain document with retrieval + format negotiation.

from langchain_community.document_loaders import WebBaseLoader

docs = WebBaseLoader(
    "https://more.md/u/ada?format=toon",
    header_template={"Accept": "text/toon"},
).load()

SimpleWebPageReader + /search endpoint replaces the per-project RAG pipeline.

from llama_index.readers.web import SimpleWebPageReader

docs = SimpleWebPageReader(html_to_text=False).load_data(
    urls=["https://more.md/u/ada?format=markdown"]
)

Call more.md as a tool from the AI SDK's tool-calling layer. Markdown in, typed object out.

import { tool } from 'ai';
import { z } from 'zod';

export const moremdLookup = tool({
  description: 'Fetch a more.md entity as Markdown',
  parameters: z.object({ handle: z.string() }),
  execute: async ({ handle }) => {
    const res = await fetch(`https://more.md/u/${handle}?format=markdown`);
    return { markdown: await res.text() };
  },
});

Register more.md as a hosted tool in the OpenAI Agents SDK runtime.

from openai_agents import Agent, WebFetchTool

agent = Agent(
    tools=[WebFetchTool(allow=["more.md"])],
)

Reference agent + FastAPI NDJSON streaming in repo `realworld-simulation/agent` (requires API keys).

pip install google-adk
# UI + compose: realworld-simulation/frontend

Microsoft AutoGen multi-agent runtime; generic HTTP tool works today, typed adapter W2.

Typed adapter ships in Wave 2. Use the framework's generic HTTP tool today.

CrewAI Task integration with payment-gate awareness. Arrives in W2.

Typed adapter ships in Wave 2. Use the framework's generic HTTP tool today.

Upsonic framework adapter; register more.md as an agent-to-agent identity.

Typed adapter ships in Wave 2. Use the framework's generic HTTP tool today.