01
01 · Capture
It reads where work already happens.
Meetings, Confluence, Jira, Google Drive, OneDrive and GitHub. Connect them once. Nobody uploads files, tags pages or writes summaries for Loom. When the work changes, the memory changes with it.
Shared memory for AI teams
Decide once. Every AI knows.
Loom turns your meetings, docs and tickets into one memory of what your company decided: who decided, when, why, and what it replaced. Cursor, Claude, Copilot and your agents all work from it. Nothing to upload. Nothing new to adopt.
Reads fromOneDriveConfluenceJiraGoogle DriveMeeting notesGitHub
Works in
Cursor
Claude
Copilot
Codex
CursorClaudeCopilotCodexOne decision changed. Every AI updated.
The problem
Ask four AIs the same question. Get four answers.
Every AI tool on your team sees a different slice of the company: whatever someone pasted, whatever it indexed, whatever it guessed. The wrong answer isn't the real cost. The real cost is the meeting to work out which answer was right, and the agent that already acted on the wrong one.
Are we still on per-user pricing?
GitHub Copilot
Yes. Pricing is €12 per user per month.
From an old draft
ChatGPT
I don't have access to your company's pricing decisions.
No access
Claude
Per user, based on the notes you shared last month.
Out of date
Any of them, with Loom
No. Per seat with a minimum of 5 seats, since 3 Oct (v3). Decided at the board meeting, replacing v2 (per seat, no minimum) and v1 (€12 per user).
Current4 sourcesOwner: Ana Ribeiro
How it works
Loom keeps the memory. Your tools keep working.
Three things happen continuously, in the background, with nothing for your team to maintain.
02
02 · Remember
It turns what was said into what was decided.
Loom pulls out decisions with their owner, date and reasons. It links them to the people, projects and systems they touch, notices when a new decision replaces an old one, and flags when two sources disagree.
03
03 · Serve
Every AI asks the same memory.
Loom connects through MCP, the open standard Cursor, Claude, Copilot and Codex already support. Assistants answer with sources. Agents act on current decisions. And each person only gets what they're allowed to see.
01 · Capture
It reads where work already happens.
Meetings, Confluence, Jira, Google Drive, OneDrive and GitHub. Connect them once. Nobody uploads files, tags pages or writes summaries for Loom. When the work changes, the memory changes with it.
02 · Remember
It turns what was said into what was decided.
Loom pulls out decisions with their owner, date and reasons. It links them to the people, projects and systems they touch, notices when a new decision replaces an old one, and flags when two sources disagree.
03 · Serve
Every AI asks the same memory.
Loom connects through MCP, the open standard Cursor, Claude, Copilot and Codex already support. Assistants answer with sources. Agents act on current decisions. And each person only gets what they're allowed to see.
What changes
Less re-explaining. Fewer surprises.
Pricing agent✓ per seat, min. 5
Support agent✓ per seat, min. 5
Docs agent✓ per seat, min. 5
Agents that pull in the same direction.
Every assistant and agent starts from the same decisions, so they stop contradicting each other and you stop re-briefing them.
Why did we drop per-user pricing?
Per seat scales for teams. Per-user did not.
Decided 12 Sep · Ana Ribeiro · 3 sources
Knowledge that doesn't walk out the door.
What your best people know is captured as they work. New joiners and handoffs start from the history, not from a hallway conversation.
Ship Q4 pricing page
Keep per-user wording in FAQ
Announce to sales
Conflicts with the 3 Oct pricing decision · ask Ana Ribeiro
Contradictions caught before they ship.
Drafting a roadmap or a customer reply? Loom flags when it conflicts with something already decided, and tells you who decided it.
12 Sep · per seat
3 Oct · minimum 5 seats
today
Why we decided, six months later.
Every decision keeps its owner, date, reasons and what it replaced. No more digging through old threads.
Security
Everyone gets answers. Nobody gets answers they shouldn't.
Loom classifies everything it reads into five tiers and checks every answer against who is asking. When a confidential decision affects your work, Loom tells you who to ask, without telling you what it says.
PublicAnyone in the companyAnswered
InternalYour teamsAnswered
RestrictedNamed groups onlyWithheld for others
EmbargoedUntil a set dateWithheld until then
PersonalOnly the personWithheld for others
Every answer leaves a receipt.
QuestionHow do we price seats now?
Sources used4 passages
Tiers appliedInternal
Withheld1 restricted passage
Policiespii-mask, tier-gate
AuditSigned and chained #a91f…3c2
Verified
Personal data maskedConnector credentials encryptedAnswers logged with sourcesTamper-evident audit trailYour data never trains models
How Loom handles your data →One memory for every AI your team uses.
Show us one workflow where your AI tools disagree. We'll show you what changes.