Cerebro

A nearly fully autonomous second brain for teams.

The "second brain" framework has existed for almost a decade, but it's predicated on human entry, editing, and verification — a fundamentally flawed system. Cerebro proposes something different: automatic capture from every team surface, custom cleaning, distillation through a memory engine, and a fun interactive UI to explore it all. No manual entry. No human-in-the-loop structuring. Your org becomes legible to AI — and to you.

Live Demo | Full Spec | Build Status

---

How it works

Six Notion Workers pull data from Slack, Gmail, Google Calendar, meeting transcripts (Granola), and Notion Docs. Each worker cleans its input through a shared Glossary normalization library and writes cleaned text into a workspace-level Short-Term Memory database in Notion.

A Hindsight Indexer Worker polls Short-Term Memory and feeds new rows into Hindsight Cloud — a memory engine that handles fact extraction, entity resolution, and observation consolidation. A Cerebro Sync Worker writes structured records into Long-Term Memory — 13 Notion databases spanning dossiers (People, Companies, Agents), actions (Projects, Tasks, Decisions), intelligence (Insights, Frameworks, Strategies, Patterns, Signals), and measurement (Objectives, Metrics).

You interact with the knowledge through a Decisions Interpreter (a Notion Custom Agent with search, detail, impact, and trend analysis tools), a force-directed graph visualization, a swipe deck feed, a HeyGen video avatar, and ElevenLabs voice chat — all powered by Claude (Anthropic API) for reasoning.

Built on real Optemization team data. Dogfooded from day one.

Architecture

flowchart LR
    subgraph Sources["Source Workers (Notion Workers SDK)"]
        Slack
        Gmail
        GCal
        Meetings
        Granola
        Docs
    end

    Glossary[(Glossary DB)]

    Sources -->|clean + write| STM[(Short-Term Memory)]
    Glossary -.->|normalize| Sources

    STM -->|poll pending| Indexer[Hindsight Indexer]
    Indexer -->|retain| HC[Hindsight Cloud]

    HC -->|webhooks| Sync[Cerebro Sync Worker]
    Sync -->|classify + upsert| LTM[(Long-Term Memory\n13 Notion DBs)]

    HC -->|reflect| Agent[Ask Cerebro Agent]
    LTM -->|read| Graph[Graph Viz]
    LTM -->|read| Feed[Swipe Deck]
    Agent --> Avatar[HeyGen Avatar]
    Agent --> Voice[ElevenLabs Voice]

Loading

What's working

• Slack Worker — deployed, ingesting messages into Short-Term Memory
• Google Worker — Gmail + Google Calendar via domain-wide delegation (built, needs deploy)
• Meetings Ingest Worker — deployed, ingesting calendar meeting transcripts into STM
• Granola Worker — ingesting Granola.so meeting recordings into STM
• Notion Docs Worker — retains docs directly to Hindsight (bypasses STM)
• Hindsight Indexer Worker — polls STM, calls retain(), flips status to indexed/failed
• Glossary Proposer Worker — scans STM bodies, extracts entity candidates, proposes Glossary rows
• Decisions Interpreter — Notion Custom Agent with search/detail/impact/trends tools
• Frontend — feed, swipe deck, and graph viz wired to 12 Long-Term Memory DBs
• Hindsight Cloud bank bootstrapped (optemization-cerebro)
• Ship workflow — branch protection, auto-merge PRs, conventional commits

See STATUS.md for the full build status.

Tech stack

Layer	Technology
App & API	Next.js (App Router), Vercel
Data workers	Notion Workers SDK
Custom Agents	Notion Custom Agents (Decisions Interpreter)
Memory engine	Hindsight Cloud (retain / recall / reflect)
Reasoning	Anthropic Claude API (Sonnet 4.6)
Video avatar	HeyGen LiveAvatar
Voice chat	ElevenLabs
Data store	Notion databases (no external DB)

Repo structure

cerebro/
  app/                    Next.js app (App Router)
    api/
      chat/               Chat endpoint (Claude reasoning)
      graph/              Hindsight graph data
      feed/               LTM feed
      deck/               Swipe deck
      ask/                Q&A endpoint
      liveavatar/         HeyGen session tokens
  lib/                    Shared libraries
    cleaning/             Glossary normalization library
    tools/                Notion Custom Agent tools
    skills/               Agent system prompts
  slack/                  Slack source worker
  google/                 Gmail + GCal source worker
  granola/                Granola meetings worker
  circleback/             Circleback meetings worker
  notion-docs/            Notion Docs source worker
  indexer/                Hindsight Indexer worker
  glossary-proposer/      Glossary entity proposer
  workers/
    meetings-ingest/      Meeting transcript ingest
    decisions-agent/      Decisions Interpreter (Custom Agent)
    documents-ingest/     Document ingest worker
    hindsight-indexer/    Hindsight indexer (alt location)
  docs/specs/cerebro.md   Full product spec
  STATUS.md               Current build status
  AGENTS.md               Agent-facing project guide

Quick start

# App (Next.js)
npm install
cp .env.example .env      # fill in DB IDs + API keys
npm run dev                # http://localhost:3000

# Any worker
cd slack/                  # or google/, indexer/, etc.
npm install
cp .env.example .env       # fill in worker-specific env
ntn workers exec <capability> --local

Team

Built at the Notion Developer Platform Hackathon (May 16-17, 2026, San Francisco) by:

• Tem (Temirlan Nugmanov) — Architecture, workers, Hindsight integration
• Kamau — Frontend, graph viz, swipe deck
• Mike — Workers, data pipeline

Optemization — a Notion consultancy for mid-market and enterprise companies.