supermemory open-sources Company Brain, a Slack bot that answers from your team's own conversations

A team's decisions, owners and answers are usually scattered across Slack threads, docs and the memory of whoever was on the call. supermemory's Company Brain, an in-Slack assistant that reads those threads and answers from them, was until a few weeks ago a paid product with thousands of users. On 2026-09-25 supermemory discontinued it and put the whole thing on GitHub under Apache 2.0.

The company-brain README calls it "a teammate in your Slack that truly knows and understands your company, and can do anything." A team installs it into a workspace and it starts reading the channels it is in. When a teammate asks a question, it answers from what has already been said, and can also open GitHub issues, write Linear tickets, or read a repo through its own tools. It can post on its own when a conversation needs something it knows.

What it runs on

Company Brain runs on the operator's own Cloudflare account. The stack, taken from the README, is Workers, Durable Objects, D1, KV and Workers AI. Memory sits on supermemory. A one-click deploy button on the repo provisions all of it and takes two secrets: a supermemory API key and one model key from Anthropic, OpenAI, Google or xAI, whichever the team already pays for. supermemory says the deploy figures out the provider from the key, and the team pays that provider directly with no markup.

Cloudflare's free plan works, and the README says the paid plan at $5 per month buys "room for the brain's full tool loop" and a built-in code sandbox in a container. On the free plan the sandbox is available through a Daytona key instead. Long, multi-step answers get cut short on the free plan by its 50 outbound calls per request.

The permissions rule

The README is specific about how memory is scoped. "Memory isn't one big bucket. It's a permissions graph, and the brain only ever reads with the asker's own access." A question asked in a public channel draws only on the shared org brain. A question in a private channel adds that channel's own memory. A DM adds the asker's personal memory and every private channel they are in. Tool writes always run under the asker's own connection, and when a request needs a tool only a teammate has connected, the brain asks that teammate with an approve or deny card. supermemory's own line: "It can't leak something you couldn't see yourself."

What is behind the release

supermemory is a real company that has been selling a memory API for AI apps, and Company Brain was one of its own products built on that API. The GitHub repository was created on 2026-09-25 and picked up 136 stars on its first day, according to the GitHub API read on 2026-09-26. Apache 2.0 means a team can fork, run it, and change what it does. The interesting part for a founder or engineering lead is that the pattern is now legible: an answer engine that reads your team's own conversations, on infrastructure you own, with the model key on your account rather than a vendor's.

Source

• supermemoryai/company-brain, Apache 2.0 license, 136 stars as read on 2026-09-26
• supermemory, the company behind the memory API