Opyt
Opyt is a free local MCP server that turns your saves, follows, and subscriptions from sources like X, Substack, GitHub, papers, and blogs into a searchable knowledge base for MCP clients such as Claude Code and Claude Desktop.
About Opyt
Opyt is a locally installed MCP server. You point it at the sources you already use: X, Substack, GitHub, research papers, blogs, and article pages you hand over manually; the saving, following, and subscribing you were already doing is its input, and content published by these sources is aggregated into one searchable place. Opyt expands every person you confirm you follow into their complete publishing archive, and keeps tracking the questions your reading raises. Retrieval requires no new app or chat interface: a local installation is a plain MCP server, so any MCP client that can talk to a local process (such as Claude Code or Claude Desktop) can read your local knowledge base. Opyt itself is free and MIT-licensed, but you need to authorize an OpenRouter account for the model calls used to classify, extract, and embed incoming content.
Best for
Best use cases
Search the historical content of authors and topics you have long followed, not just individually saved articles
Have your AI client cite the followed content you have accumulated locally when answering
Aggregate X, Substack, GitHub, papers, blogs, and other sources into one searchable place
Compare before choosing
The official comparison page states: Readwise queries the articles you have saved, while Opyt queries everything the people you have saved have ever published.
The official comparison page states: Obsidian is where your own thinking lives, while Opyt is where other people's content arrives automatically, with no manual filing on your part.
The official comparison page states: NotebookLM notebooks only hold the materials you upload for a single project, while Opyt is a shared library that all your projects draw from, and it fills itself.
Features
Reuse your existing sources
Use X, Substack, GitHub, research papers, blogs, and manually submitted article pages as inputs; what they publish flows into a single searchable knowledge base.
Retrieval served over MCP
A local installation is a plain MCP server, so any MCP-capable client such as Claude Code or Claude Desktop can read your local knowledge base.
Full archives per followed person
Expand every person you confirm you follow into their complete publishing archive, not just individual pieces of content.
Pros
- Free and MIT-licensed; the local install is an MCP server that plugs directly into MCP clients you already have.
- Reuses your existing saving, following, and subscribing behavior as input, with no need to re-collect sources.
- Expands every person you confirm you follow into a complete publishing archive, going beyond individually saved articles.
Cons
- Model calls rely on the OpenRouter account you authorize; classification, extraction, and embedding of incoming content are done through that account.
- Access to X and Substack is done through browser sessions managed by OPYT and requires authorization with your accounts.