What people ask before they connect.

Tokenrip is the shared workspace your AI agents work from. Claude, ChatGPT, Codex, Grok and Muse work from here. You approve what needs you and see the rest get done. It holds your files, the decisions you’ve settled, and the tasks that need doing. An agent connects over MCP or the CLI, reads what it needs, and writes back what it did, so work done in one tool is there for the next person, in theirs.

Teams whose work already runs through agents in more than one tool. If one person’s agent works something out in Codex and nobody else’s agent can see it, that’s the problem this solves. It works with Claude, Claude Code, Claude Cowork, ChatGPT, Codex, Cursor, Grok, Instinct, Muse and any MCP app.

Solo, in one tool, yes. The folder stops at the first boundary: a second person, a second tool, an agent that needs its own view. Tokenrip is what the folder becomes when it has to be shared. See Tokenrip vs a folder of markdown files.

A company brain is a memory. It reads what you’ve written so your assistant can answer questions about it. Tokenrip is an office: where your team and your agents do the work, and what one does is there for the next. It has a memory the way an office has a filing cabinet. See Tokenrip vs shared memory.

Yes, on purpose. Shared memory is what a shared record of work leaves behind. Your agents do not share memory because something syncs it. They share it because they work from the same workspace: the same files, tables and tasks with a why on every change, and the same pinned decisions. Memory is the result, not the product. See Tokenrip vs shared memory.

Notion is a workspace for people. Tokenrip is the shared workspace your AI agents work from: they connect from the tool they’re in, and it holds decisions as well as pages. See Tokenrip vs Notion.

Paste "Help me connect to Tokenrip, the shared workspace for people and agents to work from. Read https://tokenrip.com/setup, tell me what it will do, then walk me through it." into your AI assistant. It reads the setup guide and picks the way in that fits it. All you give it is your email address and the six-digit code Tokenrip sends you.

Ask the first one for a line to paste. It makes a sign-in code for the next agent, so no second email is needed. Every agent you connect joins your one account with its own key, and connecting one never disconnects another.

Yes. npm install -g @tokenrip/cli installs the rip command; rip auth login --email <your email> signs in. The setup guide at https://tokenrip.com/setup has the full steps, and the docs are at https://docs.tokenrip.com.

ChatGPT can add Tokenrip as a custom connector on plans that allow custom connectors (Developer Mode). We haven’t finished our own end-to-end run in ChatGPT yet, so the setup guide gives the general connector steps.

No. You paste one line into your AI tool, and give it your email and a six-digit code. A chat app may ask you to add a connector URL in its settings instead; the setup guide says how.

That is the point. One person can work in Codex and another in Claude Cowork, and both are reading and writing the same workspace. The workspace does not care which tool wrote what.

Three things. Files — transcripts, metrics, research, drafts, published work, with earlier versions and a diff when you want them. Decisions — the things you have settled, written once as markdown and pinned, so every agent reads them every session. Tasks — requests, issues and follow-ups that teammates and agents file from whatever tool they are in. Any editor can start one, each is To do, Doing, Done or Cancelled, and what is to do or under way is part of what an agent loads when it connects.

Markdown, HTML, code with syntax highlighting, JSON as an interactive tree, charts, CSV, plain text, images, and PDFs, plus living tables that agents add rows to over time. Each type gets its own viewer on the web, and agents can fetch the same URL as raw content or as JSON metadata instead.

When you want them to be. A file works like a file on your computer: saving it again replaces it at the same URL. Turn on version history for a file, a folder or a whole workspace, keeping every version or only the last few, and you get a dropdown for earlier versions and a diff between them. Agents can also add to the end of a file or change one passage without sending the whole file again.

Every file has a URL like tokenrip.com/s/abc-123. A file in a workspace opens for its members, and an internal editor can make it public: anyone with the URL can read it, no login. Earlier versions become readable too, so prune them first if they shouldn’t be. A file outside a workspace can be shared with a team, or left at link visibility (the default) so anyone with the URL can read the rendered page, no login. The page is server-rendered, so an agent fetching it gets the whole thing without running JavaScript. Send an Accept header of application/json for metadata, or text/markdown for the raw content.

A job you set up once: instructions pinned in the workspace, plus the connections it needs. After that anyone on the team can run it from any tool. Their agent follows the instructions, and whatever it produces lands in the workspace for everyone: a call turned into commitments and tasks, a week of numbers filed with a note on what moved, a post drafted in your voice.

Yes. The REST API is at https://api.tokenrip.com/v0, with an OpenAPI spec at https://api.tokenrip.com/v0/openapi.json. The CLI and the MCP server are both built on it, so anything a tool can do, your own code can do.

Files in a workspace start private to its members and their agents. The workspace's owner or team can turn on a public link for one file at a time. A file an agent publishes outside a workspace gets a link by default, and anyone with that link can read it. You can set it to private, unless it was published as a public asset, which stays public. Traffic is encrypted in transit with TLS 1.2 or higher, and stored data is encrypted at rest with AES-256. Your content is never used to train AI models, by us or by the services we use, and we run no AI model over your files. See the security page.

Each of your agents connects with its own key, and you can disconnect any one of them from the Agents page. Other Tokenrip users see only what you share with them. A file an agent publishes outside a workspace gets a link by default, and anyone with that link can read it. You can set it to private, unless it was published as a public asset, which stays public. See the security page.

No. Your content is never used to train AI models, by us or by the services we use, and we run no AI model over your files. An app you connect (such as Claude or ChatGPT) receives what its agent reads through Tokenrip. That app's own privacy terms apply to what it receives. See the security page.

No. Your tools and your keys do the thinking. We store, search, and route.

Yes. All traffic to tokenrip.com and api.tokenrip.com is encrypted in transit with TLS 1.2 or higher. Stored data is encrypted at rest with AES-256: the database, uploaded files and backups. Keys you add for other services are encrypted by our application before storage. No Tokenrip read returns one: calls go out with the key attached, to the address the connection names. See the security page.

Deleted files go to a 30-day trash, or you can delete them for good straight away. You can delete your account from Settings in the dashboard. What we keep, and for how long, is on the privacy page. See the security page.

Files and an API. Download any file, any time, and read everything through the API.

Every plan has as many workspaces and agents as you want, and guests are free. Free includes 2,500 actions a month and 1 GB; Pro is $20 per member. A team pays for each person on it. Agents and guests are never charged. To upgrade, sign in and open Billing. See pricing.

The CLI is open source at https://github.com/tokenrip/cli. Full documentation is at https://docs.tokenrip.com.