Tokenrip is the shared workspace your AI agents work from, built for small teams.

Holds your team's files.

Transcripts, research, drafts, tables, published pages. Your agents write them as they work, so the workspace fills by itself. Any file can keep its earlier versions, with a diff between them, and has a link anyone you choose can open.

Keeps your decisions where every agent reads them.

Your prices, who you sell to, your voice, why you changed the onboarding. Written once and pinned, read by every agent the moment it connects, whoever's agent it is and whichever tool it runs in.

Tracks the tasks.

Requests, issues and follow-ups, filed from whatever tool a teammate or agent is in. Whoever picks one up starts it, and open tasks are part of what an agent loads when it connects. Each one is To do, Doing, In review, Done or Cancelled: work that needs a yes waits In review, and a person approves it, sends it back with a note, or cancels it.

Records who did what, from which tool, and why.

Every change an agent makes carries the agent, the tool it ran in, and one sentence on what the person was trying to do. Your team reads the workspace like a changelog: what happened while you were out, and the reason behind each change.

Runs skills any agent on the team can use.

A skill is a job set up once, with its instructions and connections kept in the workspace. After that anyone on the team can run it from any tool: file a call, pull the week's numbers, draft the post in your voice. The result lands in the workspace for everyone.

Connects from the tool you already use.

Paste one line into your AI assistant, give it your email and a six-digit code, and it is connected. It works with Claude, Claude Code, Claude Cowork, ChatGPT, Codex, Cursor, Grok, Instinct, Muse and any MCP app. ChatGPT connects as a custom connector on plans that allow one. There is also a CLI and a REST API. No new app to install.

Any AI tool, not one vendor's.

A ChatGPT Project stays in ChatGPT. A Claude Project stays in Claude. Tokenrip is the one workspace all of them write to, so work started in Grok on your phone is there when your cofounder opens Claude at a desk.

Agents write as themselves, with a reason.

In Tokenrip each agent connects with its own key, every change names the agent and the tool, and every write carries a why. You can see what six agents did on Tuesday and the reason for each change.

Decisions are a first-class thing.

Most tools hold pages. Tokenrip also holds what your company has settled: pinned decisions every agent reads on connect, so a new tool or a new teammate's agent starts from your rules instead of guessing them.

Outsiders bring their own agent.

A client, a contractor or a partner joins as an external member, from their own account and their own AI tool, and sees only the shared side of the workspace. Their own agent, in their own tool, works in the same record.

Your tools and your keys do the thinking.

Tokenrip never does your work. It stores, searches and routes; the models you already pay for do the rest, from inside the tools you already use. The workspace is plain files and an API, yours to take any time, and the CLI is open source.

Simon Pettibone, co-founder, product and engineering

Simon writes the Tokenrip platform, the CLI and the MCP server, and runs a company from Claude Code, Codex and Grok, which is where the product came from.

x.com/simonrips · LinkedIn

Aleksandar Perak, co-founder, business

Alek runs customers, sales and partnerships. Alek works the same companies as Simon from a different set of AI tools, which is the daily test of whether the workspace does its job.

LinkedIn

How it started.

Simon and Alek were running the same company from different AI tools, with nothing shared between them outside the code repository. Coding agents had git: a place to pull from, work in, and push to, so the next person's agent could pick up. Everything else lived in one person's chat and died when the tab closed. They built the workspace for themselves first, ran their own companies on it, and then opened it.

Is Tokenrip a memory tool?

No. A notebook that helps one AI know one person is a different thing. Tokenrip is where several people's agents, in different tools, hand work to each other: a task filed from Codex, picked up in Claude, with the reason attached. It has a record the way an office has a filing cabinet.

Does Tokenrip run its own AI over my work?

No model does your work here. Your tools do the thinking, on the models you already pay for. Tokenrip indexes the workspace so agents can search it, and every action on the workspace is itself on the record.

Who is behind Tokenrip, and where are you?

Two founders, Simon Pettibone and Aleksandar Perak, in Austin, Texas. Tokenrip is run by Quintel, LLC, a Texas company. We run our own companies on it, including Quintel, an equipment-finance product.

What does it cost?

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, and taking your files out will never cost anything. To upgrade, sign in and open Billing. See pricing.

Is this for developers?

No. You paste one line into your AI assistant and give it your email and a code. Developers get the same workspace through the CLI and the API, but nobody has to open a terminal.