The memory and context engine for your agents.

Marrow Core

Most agents fail at company work because their context is out of date.

Core tracks how facts change across your email, documents and chat, flags where sources conflict, and serves each agent the current state alongside its provenance.

For
Developers building agents, and every team using Company Brain
Sources
Email, documents, chat, spreadsheets and your own systems
Runs
Hosted, customer cloud (VPC) or on-premises
Interfaces
The API, MCP, the SDKs and the CLI

Sources and facts

Core decomposes documents and messages into discrete, dated facts, so agents reason over sourced statements rather than raw text.

Traceable provenance: every extracted fact links directly to the passage it came from.

Context on request

Core assembles the exact context a task requires within the user’s permissions, and shows where each fact comes from and what it replaced.

Access boundaries hold firm: permissions are checked before retrieval, not after.

Continual learning

Core turns repeated corrections in completed work into reusable procedures for human review.

Tested before approval: each proposed procedure is benchmarked on held-out tasks first, and any approved change can be undone.

Hosting and export

Core runs hosted by Marrow, in your own cloud or on-premises, and you can export every fact, source and procedure at any time.

API, MCP and SDKs

Integrate Core through our hosted MCP server, typed TypeScript and Python SDKs, or directly via the REST API.

curl https://api.marrow.id/v1/ns/ops/search \
  -H "Authorization: Bearer $MARROW_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "When is the launch, and what has changed?",
       "include": {"related": true}}'
{
  "method": "tools/call",
  "params": {
    "name": "search",
    "arguments": {
      "namespace": "ops",
      "query": "When is the launch, and what has changed?"
    }
  }
}

Connect Core in Claude, ChatGPT, Gemini CLI or Lovable.

import { Marrow } from "@marrowid/sdk";

const client = new Marrow({ apiKey: process.env.MARROW_API_KEY! });
const context = await client.search("ops", {
  query: "When is the launch, and what has changed?",
  include: { related: true },
});
import os
from marrow import Marrow

client = Marrow(api_key=os.environ["MARROW_API_KEY"])
context = client.search(
    "ops",
    query="When is the launch, and what has changed?",
    include={"related": True})

Uniform interfaces: the same contract whether you connect through MCP, the SDKs, the CLI or the REST API.

Get access to Core

Get in touch for API access.

Contact us