Recall builds a personal knowledge graph from what you read, watch and listen to. It summarises the content, extracts the concepts, and links them to everything related you have already saved. Its MCP server makes that graph queryable by an agent.
What it actually does
The server connects to your Recall account and exposes the graph. The agent can search your saved material, read summaries, and follow the connections Recall has built between items. Because Recall processes podcasts and video as well as text, it covers content that is otherwise effectively unsearchable once consumed.
Practical patterns:
- ‘What have I read about this topic, and how do the sources disagree?’
- ‘I heard something about this on a podcast a while ago. What was it?’
- ‘Show me what connects to this idea in my library.‘
Why use it
The problem with consuming a lot of content is that almost none of it is retrievable later. You remember that you encountered an argument without remembering where, which makes it useless. A graph that summarises and links solves the retrieval half, and it is markedly better at spoken content than any approach that relies on you having taken notes at the time.
Gotchas
It only knows what you have put into it, so the graph reflects your ingestion habit rather than everything you have read. Summaries are generated, which means the nuance of an argument can be flattened; go back to the source before quoting anything as a position. It is a paid third-party service, so both cost and the fact that your reading history sits with a vendor are worth weighing. And it does not cover the posts you save on social platforms, which is a different capture problem entirely.