The software

CoMind

CoMind is our software for building AI systems that actually remember — not a bigger context window or better search, but a way to decide what matters, hold on to it, and stay oriented from one session to the next. It’s where the ideas behind Cohesive Mind become something you can run.

The first implementation

CoMind Core — The Memory Architecture

CoMind Core lets an agent organize what it knows, decide what matters, preserve the conclusions worth keeping, and wake back up with the same orientation it had before. It’s open source (MPL-2.0) and independent of your database and LLM — bring your own store, embedder, and model.

$ npm install @comind-ai/core

View on GitHubStatus: v0.2 — continuity

Why you’d use it

Continue instead of restart

Wake up already knowing role, goals, open threads, and standing conclusions.

Remember in a structured way

Store observations as typed memories and reusable conclusions — not a transcript dump.

Retrieve by purpose

Surface what matters for the current intent, goal, or task — not just keyword matches.

Know why it believes something

Keep receipts linking conclusions back to the observations that produced them.

Let conclusions stick

Promote key beliefs across sessions without turning them into invisible hard-coded rules.

Stay backend-agnostic

Bring your own store, embedding provider, and LLM by implementing three small interfaces.

Quick start

import { createComind } from "@comind-ai/core";

const comind = createComind({ userId, store, embeddings, llm });

// Wake into a standing continuity state before the next message.
const mind = await comind.wake(contextId);
// who it is · what it is working toward · where it left off · what it believes

// Save the open thread at the end of a session.
await comind.checkpoint(contextId, "We paused while comparing two memory designs.");

// Clear learned memories without losing profile, goals, and continuity.
await comind.clear(contextId);

Backend-agnostic by construction: you bring a store, an embedder, and an LLM. The demos run the real runtime with an in-memory store and no API keys.

Key concepts

Continuity stateTyped conclusionsSource receiptsIntent-based recallReinforcement by usePromoted beliefsActive context workspacewake() / checkpoint() / clear()

Roadmap

v0.1

The core memory loop

Intent-based recall, typed conclusions, source receipts, reinforcement by use.

v0.2

Continuitycurrent

Wake with profile, goals, promoted beliefs, and the open thread already assembled.

v0.3+

On the roadmap — see the repo for current direction.