# Reintersect - discuss, decide, and put agents to work.
Reintersect is where teams building ambitious things turn conversations into decisions, and decisions into work done by agents. Without standups, handoffs, and endless context dumping in prompts.
Get early access at . We’re inviting teams gradually. Start with one real project and the people working on it. Join early and help shape what Reintersect becomes.
This file is the complete text of the Reintersect homepage, including the screen reader narration of the demonstrations that play on it. Those demonstrations are the core experience, so open to watch them.
Screen reader narration of the opening demonstration: A conversation, its Decision, and the completed agent work stay together in Reintersect.
## Think together, with the details in place.
Whatever you’re working on gets its own conversation, with the people who need to see it. As you write, Reintersect finds earlier conversations about the same thing, by meaning, and shows them beside your draft.
Screen reader narration of the Discuss demonstration: Maya starts a conversation about the onboarding checklist. As she types, Sam’s earlier funnel conversation appears and she opens it before posting. Leo pushes back, Maya asks for numbers, Sam posts them, and Maya calls it.
### Readable by agents.
Agents get the whole conversation, structure and all.
Every conversation reaches an agent as a structured document: headings, code blocks, mentions, files, and who said what, in the same JSX format the SDK writes. What your team wrote is exactly what the agent reads.
### Everything happens in threads.
Replies stay attached to the conversation they answer.
Each conversation is its own thread. Replies stay in it, and you can reply to a reply, so a side question doesn’t derail the main one. Files and code go wherever they’re relevant.
### Backed by a sync engine.
The whole app feels instant, on every screen.
We spent a long time making Reintersect feel instant. Replies, reactions, and page changes show up the moment you act, on everyone’s screen. Those tiny half-second lags in chat apps that wear you down over a day aren’t a thing here.
### Find what you worked on before.
Related conversations show up while you write, intelligently found by meaning.
As you write about the checklist, Reintersect looks for conversations that mean something similar and shows them beside the composer. Open Sam’s funnel conversation, read what the team settled, and come back to a draft with nothing lost.
## Keep the decision. Skip the separate document.
Your team does the thinking. When the conversation is settled, resolve it, and Reintersect writes the Decision: what you chose, why, and who does what next.
Screen reader narration of the Decide demonstration: Maya resolves the conversation. Reintersect gathers the context and publishes the Decision to cut the checklist to three steps, with an owner on each action item.
### You decide. It writes.
Resolve it and Reintersect writes the Decision, reasons and next steps included.
Resolve the conversation when it’s ready. Reintersect gathers the replies, works out what was chosen and why, and composes the Decision with its alternatives and action items.
### Remember why you ruled it out.
Keep the alternatives and the reasons they didn’t become the plan.
The Decision lists the options the team discussed, marked chosen, rejected, or deferred, with the reason for each. Every reason links back to the reply it came from, so when someone asks why the team didn’t go wider, the answer is already written down.
### Next steps, with names attached.
Each next step has a name on it, and a link to where they said yes.
Reintersect reads the replies for commitments: someone volunteered, someone was assigned, or someone asked and got a yes. Each one becomes an action item with a name on it and a link to the reply. Nobody has to rewrite the conversation as a task list.
### Find what was decided.
Search for the outcome when the question comes up again.
When the checklist comes up again, search the workspace and the Decision comes back with the conversation it was composed from. Open it and the recorded choice, the reasoning, and the replies are on one screen.
### Update it without losing the history.
New replies produce a new version, and the earlier ones stay readable.
Sam adds a clarification about the extension, so Maya updates the Decision. Reintersect composes a new version from the new replies while the current one stays on screen for anyone reading it. The earlier version is kept, word for word.
## The reasoning stays with the decision.
Every quote in a Decision links to the reply it came from, so you can read the thinking where it happened. When something changes, the Decision updates and the earlier version stays.
Screen reader narration of the Remember demonstration: A citation opens Sam’s original funnel post. Sam adds a clarification about the browser extension, and Update composes a new version of the Decision while the existing one stays on screen.
### Back to the source.
Follow any quote in a Decision back to the reply it came from.
Click the drop-off number in the Decision and Sam’s original post opens, with the replies around it. Read the exchange as it happened, then step back to the Decision. Nobody has to reconstruct who said what.
### A Decision is a graph.
Goals, options, reasons, and actions are linked, so your agents can walk from one to the others.
Sam’s drop-off numbers, the choice to keep three steps, and Leo’s card are linked nodes in a graph. An agent picking up Leo’s card walks back to the choice and the numbers behind it.
### Bring context over MCP.
Let other agents find your conversations and Decisions.
Connect Claude, Cursor, ChatGPT, or any MCP client to the workspace. It can search across circles, list conversations, and open one with its replies and Decision. The connection is read-only: it brings the record into your tools without changing anything here.
## The whole team works with the same agent.
Put the Decision to work with an agent that already knows the conversation. Teammates keep adding to the same run, and the result comes back to the conversation it started from.
Screen reader narration of the Build together demonstration: An agent starts from the Decision. Leo and Sam add requests while it runs. The agent removes the four steps, updates the tests, exports the completion numbers, and returns a draft PR to the conversation for review.
### The plan you already have.
Agents run on the Claude or ChatGPT plan you already pay for, with no markup from us.
Each member connects their own Claude or ChatGPT account, once. Your runs use your plan and its limits, Leo’s use his, and the agent conversation stays shared with the whole team. Reintersect never resells tokens, so there is no per-token bill and no markup.
### Make it multiplayer.
Agents here are multiplayer: your whole team can work in one run at once.
Leo sets the step order. Sam flags the test fixture. Both go into the same run, with their names on them, and the agent takes them into account as it works. Anyone with access can answer its questions or send it somewhere new.
### Starts in context.
The conversation and its Decision go with the task.
Start execution from the conversation and the agent has the replies and the Decision as its brief. It also has a sense of what you’ve been working on lately, so you rarely explain where things stand. The result is posted back into the conversation.
### Works with your tools.
It connects to the apps your team already uses.
It edits code, runs tests, and opens draft pull requests, files Linear issues, reads Figma files, and connects to hundreds of apps. It goes deeper with GitHub for pull requests and Linear for a Decision’s action items. Anything it makes comes back to the conversation.
## The rest of Reintersect.
A few more things that didn’t fit above.
### Catch up in one pass.
Everything reaches you, grouped so a busy conversation is one update, not a hundred pings.
Replies on the same conversation arrive as one notification with a short summary of what changed, so twelve replies are one thing to read. Follow the conversations you care about, switch a circle to relevant activity only, or mute it for a while.
### Find it without the exact words.
Search by meaning to get back to the conversation you have in mind.
Describe what you’re looking for the way you’d say it to a teammate. Search matches meaning as well as keywords across conversations, replies, people, and circles. Agents can search this way too, and open what they find.
### Skip the recap.
Ask about a project and the agent finds the conversations and Decisions itself.
Ask why the checklist is three steps now. The agent pulls up the Decision, Sam’s funnel post, and who took which step, and cites them in its answer. You never paste the backstory into a prompt.
### Connect once for the whole team.
An admin connects an app once and every agent conversation can use it.
Connect an app at the workspace level and it becomes a tool in every member’s agent conversations. Personal connections sit alongside shared ones. Only app connections are shared: each person’s Claude or ChatGPT plan stays their own.
### Steer one agent as a team.
Everyone can line up the next request, reorder it, or send it now, while the agent works.
Sam adds the test fixture and Leo adds the step order while the agent works. Anyone with access can add an item, drag the queue into a new order, or send the next one now. Each item keeps the name of whoever added it.
### Your agent’s own cloud computer.
It can do anything you’d do on a computer, and you don’t have to be there.
Every agent gets its own computer. It can turn an image into a PDF, run a script, install tools, or check out a repository and run the tests, without your laptop being open. Whatever it makes comes back to the conversation, ready to download.
### An answer you can explore.
Agent answers can come back as charts, comparisons, and details you can open.
Ask how the new checklist is doing and the agent can answer with a chart, a side-by-side comparison, or a file to download, each with links to the conversations it drew on. Teammates react and reply in the same shared conversation.
### Start from a teammate’s conversation.
Copy a shared agent conversation and take it in your own direction.
Sam’s funnel conversation already has the numbers and the Decision in it, so Maya copies it, messages and attachments included, and asks for the changelog draft. The copy links back to the original, and Maya’s runs use Maya’s plan.
### Build something we haven’t.
A typed API and SDK, in the same JSX format agents read conversations in.
Read the workspace and write into it through a typed API, with content authored in JSX: headings, paragraphs, code, mentions, images, files. Agents read conversations in the same format, so they understand anything you publish. What you build from there is up to you.
## What a customer says
> We recently grew our team from 3 to 7. I used to spend half my day making sure everyone was on the same page. Once we moved to Reintersect, I got my full days back.
Konstantin Wohlwend, Founder & CEO of Hexclave (YC S24).
## Contact
- Email: hello@reintersect.com
- X: https://x.com/reintersect
- LinkedIn: https://linkedin.com/company/reintersect
Made by humans, together. With unreasonable care.