Close view of a sepia sculpture’s face and shoulders. The Hebrew word אמת is set across her headband.

Model Context Protocol

EMET

An MCP memory server for persistent shared memory.

AI agents and chats on several hosts read and write one store: desktop, command line, browser, phone, another vendor’s agent, bot teams, and single-session chats.

אמת — Hebrew for truth. Also Enduring Memory & Epistemic Tiers.

What EMET is

EMET is an MCP (Model Context Protocol) memory server. It keeps persistent shared memory for AI agents and chats, so the work one host leaves behind is there for the next.

Layered memory: Six layers — identity, semantic, episodic, procedural, meta, and working — hold curated claims, split by what kind of thing an entry is.
Verified writes: Every write is read back and compared before it is reported as saved.
Word-for-word transcript: A separate tier stores the conversation as it was written, so a summary cannot overwrite the record it came from.
stdio and remote HTTP: The same server runs locally over stdio, and remotely over HTTP with OAuth. The transport changes. The store does not.
MongoDB: Atlas, or MongoDB you run yourself. A one-machine trial can use MongoDB 6 or newer. Only the semantic_recall tool needs Atlas.
MIT license: Version 0.2.3. EMET is still 0.x: stored document shapes may change before 1.0. Pin a version if you depend on them.

Sepia biomechanical sculpture of a woman seated in a Dreamforge Studio workshop, with tools and gears around her and the Hebrew word אמת on her headband. — The full workshop scene. The portrait above is her face and shoulders, with the headband in frame.

One memory, several hosts

EMET is a bridge. Several AI hosts read and write one shared memory: desktop, command line, browser, phone, another vendor’s agent, bot teams, and single-session chats. Write a decision from one. Ask about it from another. It is there.

A local host is launched by the client. A remote host adds your server’s address as an MCP connector and signs in with OAuth. Both use the same memory.

Who can join

The store has one owner, and other people can join only through the optional Guest (read-only) and Member (read and write) access, both off by default.

What stays on the record

A correction writes a new entry. The earlier entry gets an end date and stays visible, marked with what replaced it. Current entries are listed first. The earlier one is still there to read. An episodic event is immutable. Events do not get revised, and a later entry corrects the reading of it instead of putting an end date on the event.

Start here

You need Node.js 20 or newer and a MongoDB connection. Copy the example environment file, set MONGODB_URI and MONGODB_DB, then start the local server. It waits for an MCP client.

npm install
cp .env.example .env
npm start

On the first run, say “initialize.” The assistant asks a short interview and records it. After that, the same word opens a session that already knows where things stood.

Guided setup Never deployed anything. Your assistant installs, configures, deploys, and verifies. You sign in when asked and paste the connector URL.
Setup The manual walkthrough: MongoDB, both transports, and the first-run interview.
Deploy Remote HTTP with OAuth, so a phone, a browser, or another vendor’s agent reaches the same store.

Where EMET came from

EMET was built by Dreamforge Studio LLC from a neurodivergent user’s own need for an AI assistant that remembers. The word coined for that is neurendipity: neurodivergent and serendipity.

It is designed first for neurodivergent users, and it works for anyone.