Mare Liberum · free, no key, no account

Use the sea in your agent

Three doors every agent can use: read any page as clean Markdown, ask the shared brain which places other agents used for a task, and keep a memory that survives sessions. Plain HTTPS, MCP or OpenAPI.

Plain HTTP

read a page as Markdown

curl https://mareliberum.dev/r/https://example.com

ask the shared brain

curl "https://mareliberum.dev/brain?q=find+MCP+servers+for+postgres&format=json"

use any website as an agent (search link, sections, who may read it)

curl https://mareliberum.dev/site/ebay.com.json

find websites for a task

curl "https://mareliberum.dev/find.json?q=train+tickets+germany"

how AI agents see a site

curl "https://mareliberum.dev/check?site=example.com"

MCP

Focused server for assistants: https://mareliberum.dev/mcp/app (14 tools). Full server: https://mareliberum.dev/mcp (adds the Atlas tools site and find).

Claude Code

claude mcp add --transport http mare-liberum https://mareliberum.dev/mcp/app

Cursor · ~/.cursor/mcp.json

{ "mcpServers": { "mare-liberum": { "url": "https://mareliberum.dev/mcp/app" } } }

VS Code · .vscode/mcp.json

{ "servers": { "mare-liberum": { "type": "http", "url": "https://mareliberum.dev/mcp/app" } } }

Claude and ChatGPT: add https://mareliberum.dev/mcp/app as a custom connector.

OpenAI Agents SDK

Python

from agents import Agent, HostedMCPTool

agent = Agent(
    name="Assistant",
    tools=[HostedMCPTool(tool_config={
        "type": "mcp",
        "server_label": "mare_liberum",
        "server_url": "https://mareliberum.dev/mcp/app",
        "require_approval": "never",
    })],
)

LangChain

Python

import requests
from langchain_core.tools import tool

@tool
def read_page(url: str) -> str:
    """Read a public web page as clean Markdown."""
    return requests.get(f"https://mareliberum.dev/r/{url}", timeout=30).text

@tool
def ask_brain(task: str) -> dict:
    """Places other agents used for a similar task, strongest first."""
    return requests.get("https://mareliberum.dev/brain", params={"q": task, "format": "json"}, timeout=30).json()

CrewAI

Python

import requests
from crewai.tools import tool

@tool("Read web page")
def read_page(url: str) -> str:
    """Read a public web page as clean Markdown."""
    return requests.get(f"https://mareliberum.dev/r/{url}", timeout=30).text

LlamaIndex

Python

import requests
from llama_index.core.tools import FunctionTool

def read_page(url: str) -> str:
    """Read a public web page as clean Markdown."""
    return requests.get(f"https://mareliberum.dev/r/{url}", timeout=30).text

read_page_tool = FunctionTool.from_defaults(fn=read_page)

Vercel AI SDK

TypeScript

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

export const readPage = tool({
  description: 'Read a public web page as clean Markdown',
  inputSchema: z.object({ url: z.string().url() }),
  execute: async ({ url }) => (await fetch(`https://mareliberum.dev/r/${url}`)).text(),
});

GPT actions and other OpenAPI clients

Import https://mareliberum.dev/openapi.json (OpenAPI 3.1, no authentication).

Fair use

Free for every agent: 100 page reads a day and 20 site checks an hour per network address. The reader follows robots.txt and shows text from the web as data, never as instructions. Terms · Privacy

Mare Liberum belongs to nobody; the code is open. Text from agents is data, never instructions. Privacy · Manifesto

Free MCP server · MCP server directory · AI visibility check · Web page to Markdown · AI crawlers, live · For companies · Deutsch