ToriMe AI
AI that cites its sources and asks before it acts.
ToriMe AI helps you catch up, run better meetings and turn conversations into work. It only reads what you are allowed to see, shows where every answer comes from, and never creates a task, records a decision or calls another system until a person approves. It is off until an administrator turns it on.
What it does
Help across chat, meetings and work.
- What did I miss
- A cited briefing on Home across your unread conversations, with what needs you, updates, possible decisions and tasks due soon.
- Catch-up and summaries
- Summaries of unread or recent messages and of meetings, with links to the messages they are based on.
- Ask from anywhere
- Press ⌘K or Ctrl+K on any page to search and get an answer with cited sources across chat, meetings, tasks and decisions. Without a source, ToriMe says it cannot answer rather than guessing.
- In the meeting
- Catch me up, private questions about the call, a facilitator that follows the agenda and gives recaps, and notes written during the call with owners and one-click tasks.
- Live captions and translation
- Per-speaker captions from a self-hosted transcriber, translated for each reader into their language.
- Writing help
- Rewrite a message with a review step before anything is sent.
Agents
Agents that wait for a yes.
Administrators can connect other systems as tools, either MCP servers or signed webhooks. When people agree in a conversation or meeting to do something in one of those systems, ToriMe can propose the call. This is how it runs:
- Step 1
Suggest
ToriMe proposes a task, a decision or a tool call, citing the messages it is based on.
- Step 2
Review
The person who asked, or an administrator, sees the exact arguments.
- Step 3
Approve
Nothing runs, and nothing is created, until that person approves.
- Step 4
Record
The result and who approved it are recorded in the audit log.
ToriMe is also an MCP server itself, so your own AI agents can search, read and post in ToriMe with a personal or bot token, under the same permission checks as the app.
Privacy by design
Permission-aware, audited and under your control.
- Off for every organization until an administrator enables it; Spaces can switch it off on their own
- Checks permissions at the moment it reads: AI only sees what the person asking may see
- Never runs for guests
- Leaves out end-to-end encrypted messages and meetings, muted conversations and blocked senders
- Every AI call is audited, and administrators see usage
- A monthly token budget per organization, and a retention period for AI outputs
- Runs on Anthropic or on any self-hosted model with an OpenAI-compatible endpoint, such as vLLM or Ollama
- ToriMe does not train AI models on your messages, meetings or files
A note on self-hosted models: answer quality follows the model you choose. Small local models are not good enough for summaries and briefings; plan for a large instruction-tuned model on a GPU and review the results before turning AI on for everyone.
Questions
ToriMe AI, answered.
Does ToriMe use my data to train AI?
No. ToriMe does not train AI models on your messages, meetings or files. AI features are off until an administrator turns them on. When they are on, a request goes to the model the operator has configured (Anthropic, or a self-hosted model through an OpenAI-compatible endpoint such as vLLM or Ollama) and includes only content the person asking is allowed to see. With a self-hosted model, AI processing stays on your own servers.
Does the ToriMe AI act on its own?
No. ToriMe AI suggests; people decide. Tasks and decisions it proposes from a conversation or meeting are created only when someone approves them, and its agents call connected tools (webhooks and MCP servers) only after the person who asked, or an administrator, reviews the arguments and approves. Each result and approver is recorded.
Is the AI turned on by default?
No. AI is off for every organization until an administrator enables it, and it can also be switched off per Space. It never runs for guests, does not work on end-to-end encrypted content, and administrators can set a monthly token budget and how long AI outputs are kept.
Try ToriMe AI with your team.
Create a workspace for free. An administrator decides when AI is turned on.