Keep the thread.
Weave everyone in.

An agent-native meeting app for thinking independently and contributing together. Give every idea a way into the conversation—with private voice support, Codex, and a shared whiteboard.

See how the agents work together

Illustrative conversation · every voice has a thread

AdaCan we ship the new flow on Friday?
MikaThe happy path looks ready.
YouI have a concern about failed payments.
JunLet’s look at that before we decide.

Good ideas don’t always get a turn.

Keep critical thinking in the conversation, while it can still shape the decision.

A proposal lands. Everyone agrees before the doubts are heard. That’s the groupthink risk we care about.

Or the conversation moves so quickly that someone waits to speak—and loses their thought along the way.

Weave In gives those ideas a place to be captured, worked through, and brought back to the room. Right inside the meeting, before the decision is made.

Three agents.
Room for your way of thinking.

Shared focus, private conversation, and work that moves forward alongside the meeting.

Omni

For the whole room

Bring an unanswered concern back into view, or return to an explicit meeting goal. Omni prepares a question silently and speaks after someone approves. Your team decides where to go next.

Muse

For thinking out loud, privately

Talk through a half-formed idea or ask about a shared screen without interrupting the room. Powered by GPT-Live-1, Muse gives you a real-time voice side chat. You choose what to share.

Codex

For work that takes a little longer

Bring the personal agent you already use. Ask it to check code, research a question, compare approaches, or build a diagram while the conversation continues. Share the result when it’s ready.

Your meeting.
Inside your agent’s browser.

Weave In works in a regular browser. In the Codex built-in browser, your agent can work directly with the room.

WebMCP gives Codex structured tools to read the live transcript, public chat, shared screen, and whiteboard within the access you allow.

Ask it to investigate a question, then post its findings to chat or update the board. The context comes from the meeting. The work comes back to everyone.

Meeting contextWebMCPCodex

Findings to chat. Editable diagrams to the board.

One board.
Everyone can build on it.

Keep a process, an architecture, or an open question in view. Point to it, challenge it, change it—together.

We keep the whiteboard inside our meeting environment so people and agents can work on the same thing. Through structured tools, Codex reads the actual nodes, labels, and connections, then edits specific elements.

Supported Mermaid diagrams become editable whiteboard elements. You adjust them visually. Your agent reads the updated structure and picks up where you left off.

Illustrative workflowCodex creates the structure. Select “You refine” to see a human edit.

From a quiet concern
to a better discussion.

A feature review, with space for everyone’s contribution.

Work it through with Muse.
You spot a payment edge case. A private conversation helps you turn the concern into a clear question.
Give Codex the investigation.
It checks the implementation and, at your request, shares its findings and a flowchart with the room.
Refine the answer together.
The team edits the board. Omni brings the open agenda question back before everyone commits.

Where your data goes

Clear boundaries for a shared room and your private thinking.

Between participants: Video, voice, screen share, chat, files, whiteboard edits, and captions travel over encrypted WebRTC connections between participants. Connections use a direct path when possible; Cloudflare TURN can relay encrypted packets when the network requires it. The relay handles connection metadata such as IP addresses and timing.
To AI providers: Your microphone goes directly to your caption provider. Your browser sends permitted assistant context, conversation and tool results directly to the configured AI provider; private assistant conversations stay out of the public room.
Our server: Our signaling Worker connects browsers, issues short-lived caption and TURN credentials, and initializes assistants. The room service stores assistant settings and coordination state. Automatic reminders send recent transcript text and previous reminder evidence through the Worker to an AI provider.
In this browser: Your display name and caption preferences stay in local browser storage. A bounded checkpoint in this tab keeps text chat, transcripts, reminders, and private Muse conversation for refresh or rejoin, valid for 12 hours after the last save. Leaving keeps that checkpoint. File bytes and the whiteboard stay in memory, so export before everyone leaves. AI-provider retention is separate from this browser storage.

Bring your next idea.

Video, screen sharing, live captions, and a shared board. Up to eight people. One link. No account needed.

Create a room