Give OpenClaw Real Memory: Cognee + Slack + Notion Knowledge Graph Pipeline
What if your AI agent actually knew your company? Not just the doc you pasted into the chat but every Slack thread, every Notion page, every postmortem, every owner, and every decision made along the way?
In this video, Brandon will show you how to build a fully automated pipeline that ingests Slack and Notion into Cognee then plugs it into OpenClaw (the open source AI agent with 379,000+ GitHub stars) so your agent finally has real, connected memory.
Out of the box, OpenClaw's memory is just markdown files. That works until you ask a question whose answer is spread across multiple documents like who owns which ticket, which incident caused which policy change. Those answers live in the connections between files. That's exactly what Cognee's knowledge graph generates.
At the end, Brandon asks the agent a single question: "What happened in the March incident, what did we change because of it, and who owns each follow-up?" and it pulls the full answer from Slack conversations, Notion postmortems, architecture decision records, and sprint retros, with every owner correctly attributed.
🔗 Chapters:
00:00 The result: an agent that knows your whole company
01:34 The architecture: SlaCrawl + NotCrawl + Cognee + OpenClaw
02:34 Creating a read-only Slack bot ("Memory Crawler")
03:05 ⚠️ The exact OAuth scopes SlaCrawl needs (bot + user)
04:28 Inviting the bot to channels
04:55 Installing and running SlaCrawl
06:31 Syncing Slack into a local SQLite database
07:25 Python script to convert SQLite → Markdown
09:33 Installing NotCrawl for Notion
10:00 Creating a Notion integration (read-only)
11:11 Syncing Notion and exporting to Markdown
12:09 Installing OpenClaw
13:18 Sanity-checking the agent
13:49 Running the Cognee plugin in Docker
14:48 Installing and configuring the Cognee plugin for OpenClaw
17:22 Restricting allowed plugins for security
17:28 Copying Slack + Notion Markdown into OpenClaw's memory
19:12 Verifying ingestion via logs
20:02 Asking the agent a real cross-source question
20:46 The answer — Slack + Notion stitched together
22:02 Recap
🛠️ Tools used:
Cognee (with the Cognee plugin for OpenClaw)
OpenClaw (open source AI agent)
SlaCrawl — Slack crawler from the OpenClaw ecosystem
NotCrawl — Notion crawler from the OpenClaw ecosystem
Docker (to run the local Cognee server)
Python 3 (only the standard library — no extra packages required)
OpenAI API (or swap in Ollama to keep everything 100% local)
🔒 Local-first by design: Slack and Notion data lands on your machine first, gets mirrored to Markdown, and Cognee builds the graph in a local Docker container. The only thing leaving your machine (in this demo) is LLM calls to OpenAI and you can keep even that local by choosing Ollama during OpenClaw setup.
▶️ Catch up on the rest of the series:
Part 1 — Cognee SDK: • Build a Company Brain in Python with Just ...
Part 2 — Cognee Cloud: • Share Your Company Brain with the Whole Te...
Part 3 — AWS Bedrock with LiteLLM: • Connect Cognee to AWS Bedrock — Keep Your ...
Part 4 — Ingesting Slack + Notion: • Connect Slack + Notion to Cognee — Build a...
Part 5 — Running Cognee 100% locally with Ollama: • Build a Knowledge Graph on Your Laptop wit...
The result: an OpenClaw agent that recalls your team's actual history across every tool your team uses.
👍 Like and subscribe for more videos on Cognee, knowledge graphs, AI agents, and local-first developer tooling.
#OpenClaw #Cognee #KnowledgeGraph #AIAgents #Slack #Notion #Python #Docker #LLM #AI #DeveloperTools