wikipethia

What's in it

59,000+ documents across 10 sources, 2017 to present, as of October 2026. Everything comes from Ethereum's own forums, repositories, and blogs. Every source is declared in sources.toml.

Missing a source? Suggest one.

How it answers

Search is hybrid, in three layers:

• Ranked lexical search (FTS5 + BM25). Rare terms outweigh common ones, title and author hits outweigh body mentions, and stemming matches "exits" to "exit". Exact tokens like EIP-4844 or an author name hit precisely.
• Semantic search (embeddings). Every chunk is mapped by a small local model to a point where meaning, not spelling, determines distance. A question about "PBS" finds proposer/builder-separation posts that never use the acronym.
• Hybrid fusion (reciprocal rank fusion). Documents strong in either list surface, and exact-term hits are never diluted by the vector side.

Every result carries a stable doc id, author, published date, URL, and its source's tier. Ethereum research supersedes itself, and a 2019 post can flatly contradict a 2024 one, so the server tells the model to weigh dates when sources disagree.

Connect

A hosted, read-only endpoint. No account, no key. It is rate limited and run non-commercially.

https://mcp.wikipethia.org/mcp

It serves the same corpus the downloadable releases give you, and follows new releases on its own: a published snapshot is live within about 15 minutes.

Claude (claude.ai)

1. Settings → Connectors → Add custom connector
2. Paste the URL above. No authentication.
3. Enable it from the chat's + menu and ask Ethereum questions.

On Team and Enterprise workspaces only an organization owner can add connectors; members then just click Connect.

ChatGPT (chatgpt.com)

1. Settings → enable Developer mode (beta)
2. Add an MCP server with the URL above. Authentication: none.
3. Pick wikipethia from the + menu in a chat.

Needs Plus/Pro or above.

Claude Code

claude mcp add --transport http wikipethia https://mcp.wikipethia.org/mcp

Other clients

Any client that speaks MCP's streamable HTTP transport can use the URL directly. The usual config shape:

{
  "mcpServers": {
    "wikipethia": { "url": "https://mcp.wikipethia.org/mcp" }
  }
}

claude.ai, ChatGPT, and Claude Code have been tested against the endpoint. Other clients should work but have not been.

Run it yourself

The hosted endpoint is a convenience. The corpus is a single SQLite file, published as a corpus-* release, and the server is one Rust binary.

• Locally: download the latest release and serve it over stdio to Claude Code or any MCP client. The README quickstart is a few commands and needs about 1.5GB of disk.
• From the sources: wikipethia build crawls everything itself, politely (one request per second to the forums) and resumably. Expect several hours.
• Your own endpoint: deploy/ holds the systemd units, nginx config, and runbook this one runs on.

Licensing

The corpus is not offered under a single license: each document stays under its source's own license, carried in the corpus metadata.

In short: for non-commercial use (research, learning, personal projects) the whole corpus is available, with credit. For commercial use, everything except the two forums (ethresear.ch and Ethereum Magicians) is available. Every document already carries its author, date, and URL, so attribution comes for free, and every result carries its source, so the corpus can be filtered to whichever sources suit your needs. The README has the full statement.