Free Search MCP server

free-search-mcp

<!-- mcp-name: io.github.sweetcornna/free-search-mcp -->

free-search-mcp is a local-first Model Context Protocol server that needs no API key. It lets any LLM (Claude, GPT, a local Ollama model, …) search the web, fetch and clean up pages, and read documents, and you never sign up for a search API.

It combines ideas from several open-source MCP servers in one Python package, and adds the output shaping for LLMs and the reliability work that each of them lacked.

research("how does reciprocal rank fusion work", depth=3)
   ↓
# Research brief: how does reciprocal rank fusion work
_engines: duckduckgo, bing, anysearch · sources: 3 · ~3,400 tokens_

## Sources
- [1] Reciprocal rank fusion | Elasticsearch Reference — <https://…>
- [2] Hybrid Search Scoring (RRF) | Microsoft Learn — <https://…>
- [3] RRF explained in 4 mins — Medium — <https://…>

## Documents
…full Markdown bodies of each page, ready for the LLM to read…

That was one tool call. It returned three sources with their full text, and no API key was involved.

Quick start: install the plugin (recommended)

You do not need to sign up, get a key or clone the repository. The only prerequisite is uv.

In Claude Code, inside a session:

/plugin marketplace add sweetcornna/free-search-mcp
/plugin install free-search@free-search-mcp

From a shell it is the same two steps:

claude plugin marketplace add sweetcornna/free-search-mcp
claude plugin install free-search@free-search-mcp -s user

In Codex:

codex plugin marketplace add sweetcornna/free-search-mcp
codex plugin add free-search@free-search-mcp

Codex pitfall. If ~/.codex/config.toml already has a

[mcp_servers.search] entry from an earlier codex mcp add, it shadows the

plugin's server of the same name. Codex gives no warning and keeps running

the old one. Remove the entry first: codex mcp remove search.

The plugin ships 11 MCP tools, a source-verification skill and a small quick-search agent. The MCP server search is pinned to the plugin's own version. The skill verified-research loads on demand and tells the agent when a search snippet is not enough and how to check a date. In Claude Code, the agent free-search:quick-search answers one lookup in its own context on haiku and hands back a few sentences with dated source URLs. The server and the skill load in both hosts: Codex lists the skill as free-search:verified-research (checked with Codex CLI 0.154). Codex and other hosts get the same agent as a file or a prompt, and an optional ask tool lets the server do the delegating. Delegating a lookup covers every route. /plugin update free-search moves all of it to the next release (in Codex, run codex plugin marketplace upgrade and then codex plugin add again). The plugin section under Install has the details and the token cost.

Nothing else is needed. Every engine that a search reaches on its own works without an API key, and the first run downloads the package from PyPI.

For Claude Desktop, other MCP clients, a source checkout or Docker, see Install.

Optionally, install Chromium once. The browser-rendered engines (brave, startpage, zhihu, …) and JS-heavy page fetches need it:

uvx --from free-search-mcp playwright install chromium

Shortened. The full README is on GitHub.