---
title: Engram — portable agent memory
nav: What Engram is
description: Versioned, permissioned context in a Postgres you own, reachable by any MCP client behind any model.
section: Engram
order: 20
---

Models are interchangeable. Your context is not. Engram is where an organisation's durable
knowledge lives — versioned, permissioned, searched by several engines at once, and
reachable by any MCP client behind any model. Every document, embedding and revision is a
row in a Postgres you control.

## Why it exists

A model session ends and takes everything with it. The next tool starts from nothing, so the
same context gets re-pasted, re-explained and re-discovered — by every agent, every time.
Engram is the layer that stops that: write once, and every application you run can use it.

## What makes it different from a vector store

**Portable.** Reachable by any MCP client, from any harness, behind any model. Context
written by one agent is available to the next — including the one you have not chosen yet.
Changing model does not cost you what you knew.

**Persistent and versioned.** Objects carry revision history, so you can see what an agent
knew and when it changed. Knowledge accumulates as an asset instead of evaporating at the
end of a session.

**Rich, not just a vector blob.** Hybrid retrieval — semantic and keyword together — over a
typed object graph, so relationships like *supersedes* or *references* are explicit and
traversable. Engram also links similar objects by itself, and marks those links as its own, so
you can always tell a relationship somebody asserted from one it inferred.

**Permissioned and yours.** Scoped by group and namespace, so context is shared deliberately
rather than pooled by accident — and all of it in a database you own.

## Bring your own database

Point Engram at your own Postgres and **your content lives there**: namespaces, objects,
chunks, embeddings and their history are rows in your database. Read them with SQL whenever
you like. Back them up with the tools you already trust. Leave, and the content stays where
it always was — with you.

Requirements: Postgres with the **pgvector** extension, which Engram enables for you at
registration along with `uuid-ossp`. Embeddings are stored as `vector(1536)`. Postgres 16 is
what we develop and test against.

> **One honest exception.** The **library registry** — which libraries exist and who belongs
> to them — is read from the OneDroid platform database, not from yours. Your knowledge is
> sovereign; the membership index that points at it is not. If a claim of total sovereignty
> matters to your compliance case, this is the line, and we would rather you heard it here
> than discovered it in an audit.

> **Schema migrations run when you register**, not continuously. A database registered a
> while ago does not automatically receive later schema changes — re-registering is
> idempotent and brings it current.

The same model serves a single developer's notebook and a regulated enterprise knowledge
base.

## Connecting

Engram has its own endpoint:

```
https://engram.onedroid.ai/mcp
```

It is also available through Synapse as a connection, which is how you get it governed and
audited alongside every other tool in the hub. One account covers both — free for individuals
and small teams.

In the official MCP registry it is published as `ai.onedroid/engram`.

## Where to go next

- **[Using Engram](/engram-usage)** — mint a token, connect a client, write, search, read the
  scores, share with a team, and the known issues.
- **[The Engram web app](/engram-web-app)** — every screen, with screenshots: libraries,
  namespaces, objects, web links, tokens, sharing, embedding, and bringing your own database.
- **[Engram tool reference](/engram-tools)** — all 45 `engram_*` tools and how they behave.