# corpus MCP server

> Enables semantic search and structured metadata queries over personal documents, emails, and records using pgvector and OpenAI-compatible embeddings.

- Repository: https://github.com/ryanjdillon/corpus
- Page: https://hostedmcp.org/servers/ryanjdillon/corpus
- Transport: —
- Stars: 0
- License: MPL-2.0

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `CORPUS_DATABASE_URL` | no | no |  |
| `CORPUS_OPENAI_API_BASE` | no | no |  |
| `CORPUS_OPENAI_API_KEY` | no | no |  |

## FAQ

### What kind of questions can I ask corpus?

You can perform semantic searches to find items by meaning, structured queries to filter by specific metadata facts, or query enriched priority signals like upcoming actions.

### Can I run this locally for privacy?

Yes, you can run a local embedding model so that no data leaves your network.

### How does corpus handle data security and access control?

It uses configurable storage tiers that allow you to define different access rules, such as providing raw data to trusted callers and sanitized, non-sensitive data to cloud-model agents.
