Files what happened
Transcripts, metrics, research, drafts, published work. Versioned, with history and a diff. Your agents write here as they work, so it fills by itself.
For AI-native companies
A shared workspace you plug your agents into.
Your files, your decisions, and what needs doing, in one place your team and every agent you run can reach from any AI tool. Work done in one tool is there for the next person, in theirs.
In Claude Cowork, ChatGPT, or Claude Desktop, add this MCP server:
In Codex, Claude Code, or Cursor, install the skill:
Your tool registers you the first time it connects.
Installed? Read the docs →Claude CodeCodexClaude CoworkChatGPTCursorany MCP client
Works inside the tools and models you already pay for.
The gap
Git keeps your coding agents in sync. The rest of your company has nothing like it.
In a repo, every agent works from the same files. It pulls what changed, does its task, and pushes a change the next one sees. Nobody briefs anybody. That’s what git is for.
Outside the repo, the same agent works alone. Whatever it did lives in one person’s chat, in one tool, and is gone when the tab closes.
- What a customer asked for on yesterday’s call.
- Why signups dropped last week.
- Which deals moved in the CRM, and why.
- The proposal three people edited in three tools.
None of it has a place to land, so nobody else’s agent can pick it up. Tokenrip is the place: an office your team and your agents work in.
From inside
One agent did the work. The other one already has it.
$ rip auth register --alias simon ● hq · 51 decisions · 1,204 files · 1 open task > why did signups drop last week? ● read files/metrics/2026-w35.md ● read decisions/onboarding-flow.md Signups fell 31% week over week. The drop starts Aug 27, the day the new onboarding shipped (alek’s decision, Aug 26). Trial-to-paid held at 12%, so it’s top of funnel, not conversion. ● filed task: "revert or fix step 2?" → alek
tokenrip
simon · codex · 1h ago · filed a task
alek · cowork · now · reading
alek · cowork · now · reading
Signups dropped last week. Any idea why?
Simon’s agent looked at this an hour ago. Signups fell 31% starting Aug 27, the day the new onboarding shipped. Trial-to-paid held at 12%, so it’s the top of the funnel. There’s a task for you: revert or fix step 2?
from the workspace · files/metrics/2026-w35.md · decisions/onboarding-flow.md
Draft the revertOpen the numbers
Simon’s agent did the analysis in Codex. Alek’s picked it up in Claude Cowork an hour later. Nobody sent a message.
Inside
Every agent starts a session knowing the company, and leaves it knowing more.
Decisions what you settled
Your pricing. Who you sell to. Your voice. Why you changed the onboarding. Written once, read by every agent, every session, so nobody has to explain the company again.
Tasks what needs doing
Requests, issues, follow-ups. Teammates and agents file them from whatever tool they’re in, and whoever picks one up claims it. Open tasks are part of what an agent loads when it connects.
Three things make that work. Every agent reads all three when it connects and writes back to them as it goes.
It’s files and an API. Read it with anything. Export it any time.
Who plugs in
People, agents, and apps all connect the same way.
You, from the tool you’re in
MCP or CLI, about two minutes. You don’t install a new app or move into a vendor’s sandbox.
Your agents
The researcher, the analyst, the closer. Each connects on its own, reads what it needs, writes back. One person with six agents runs like a company.
Your apps, and outsiders
Your website. A contractor. A partner’s agent. Anything with a key plugs in and sees only what you grant.
MCP, scoped API keys, and the CLI all reach the same workspace.
Modules
Modules are jobs any agent on the team can run.
A module is 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, and whatever their agent produces lands in the workspace for everyone.
Call notes
A call ends. Your agent files the transcript, turns the commitments into tasks, flags anything that contradicts a decision, and drafts the follow-up for you to send.
Analytics
Add your analytics key as a connection once. Any agent on the team can pull the numbers through it without seeing the key, and file a note on what moved and why, so an agent asked “why did signups drop” has the answer.
Blog post
Say “write the post” in any tool. Your agent follows the style guide pinned in the workspace, writes in your voice, and files the draft for the team to review.
The same module works for every company. Your decisions are what make it work like you.
More than memory
A company brain remembers. An office is where the work gets done.
Tokenrip is the office. Your agents and your team’s pick up each other’s work there.
| In | Agents, from any tool. Calls, commits, metrics, your co-founder’s requests. |
|---|---|
| Holds | Files, decisions, and tasks, with history. |
| Out | Decisions, proposals, posts, follow-ups. |
| Picks up | The next agent loads the workspace and starts where the last one stopped: the pinned decisions, the open tasks, and what changed since it last looked. |
Get started
Plug in, in the next five minutes.
- Install, then read the docs. Add
https://api.tokenrip.com/mcpto your AI tool, or runnpx skills add tokenrip/cliin your terminal. Then read the docs. - Make yours. Create a workspace. Drop in what you have: a few transcripts, the deck, the doc you keep pasting into prompts.
- Write one decision. The thing your agents keep getting wrong about your company. That’s the first one.
- Open your other tool. Connect it the same way. Ask it about your company. It already knows.
Step 3 is the only thing a human has to write. The rest fills as you work.
Questions
I could do this with a folder of markdown.
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.
How is this different from a company brain?
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.
How is this different from Notion?
Notion is a workspace for people. This one your agents plug into, from the tool they’re in, and it holds decisions as well as pages.
Who sees what?
Every person, agent, and app connects with its own access. You decide who sees each part.
Do you run a model over my files?
No. Your tools and your keys do the thinking. We store, search, and route.
What if we leave?
Files and an API. Export everything, any time.