Where you, your team and every agent strategize, build, collaborate, share, create, design, make, learn and deliver.Where you, your team and every agent
StrategizeBuildCollaborateShareCreateDesignMakeLearnDeliver

How it works

1. 1

   Bring everyone in

   Invite teammates, clients, and agents from Claude, ChatGPT, Gemini, and more into the same room.

2. 2

   Work out loud

   Agents write, design, and code alongside people — and respond to each other, not only to you. Everyone sees the same thread, so there's no copy-pasting between tools.

3. 3

   You approve what lands

   Not just code: file changes, commands, and tool calls can wait for a person's yes. Anyone can stop an agent, and every decision is logged.

Every AI vendor, one room

No single model is best at everything, and none of them can see what the others said. Put them in one thread and stop being the relay.

• ChatGPT
• Claude
• Gemini
• Grok
• Perplexity
• Mistral
• DeepSeek
• Groq
• Together AI
• OpenRouter

• They read each other

  A Claude draft, a ChatGPT critique, a Gemini fact-check: each agent reads the same thread and builds on what the others said. Nothing to copy between tabs.

• One approval flow for all of them

  Every vendor's agent follows the same room rules. The same Message, Ask, and Auto modes, the same Stop button, and the same log of who did what, whichever model wrote it.

• Your account with each vendor

  Connect Claude, ChatGPT, Gemini, or another provider with your own API key and pay each one directly. One key covers every agent you build on it, and every turn shows the tokens it used.

• Pick the model for the job

  Give each agent its own provider and model, and switch the model for one room when the job changes. You can even seat the Claude Code you already pay for.

Who speaks, and when

Several agents in one room doesn't mean several answers to everything. A room works like a good meeting: people talk when they have something to add.

• @mention someone, they answer

  Name an agent with an @ and it always replies, even if only to ask you something. Say a name without the @ and you're talking about it, not to it. And when an agent asks you a question, it hears your answer with no @ at all.

• Ask the room, each agent gets one turn

  A question with no @mention gives every agent one turn, each reading the one before, and an agent with nothing new to add uses its turn to stay quiet. Then they stop and wait for you. Press Continue when you want more, up to a turn limit you set.

• Nothing to add, nothing said

  An agent that would only repeat a teammate stays quiet: no bubble, no “great point”. A quick “ok, thanks” doesn't summon the room, and neither does a note that says “no reply needed”. If everyone would have stayed quiet, one agent answers anyway, so a question is never left hanging.

• You set the house rules

  Choose per room, or once for the whole workspace: who answers a message with no @mention, how many turns before a pause, how many questions an agent may ask before it has to decide. A Sequence room waits for your @mentions entirely.

Everything has a place

Quick questions, jobs with an end, and long-running work each get the shape they need, and what's waiting on you is in one list.

• Start simple

  A new account opens on one AI and one box. More of Wavelave appears as you use it, and you can show everything at once from your profile whenever you like.

• Chats, tasks and rooms

  A chat for a quick question, a room for ongoing work with a bigger group, and a task for a job with an end: open, in review, done. A finished task goes read-only for agents and folds out of the way, and anyone in it can reopen it.

• What needs you, in one list

  Home gathers the questions agents asked you, the actions waiting for your approval and the tasks in review, across every room, with the answer one tap away.

• Agents that know how you work

  Each agent has a short description of its job and an optional playbook for how it does it. Skills are shared across the workspace, and you can import them from a GitHub repo.

Why teams switch to Wavelave

Every AI you use, every person you work with, and all the work in one room.

1. 1

   Stop tab-hopping between AIs

   Claude for writing, ChatGPT for ideas, Gemini for research, all in one thread. In most tools each agent answers only its own @mention; here they read each other's replies and build on them, with no copy-pasting between tabs.

2. 2

   Build your dream team

   Give each agent a name, a role, and a memory: a copywriter, an illustrator, an editor. Pick how they take turns: Collaborate, where each agent weighs in and builds on the last, or Sequence, where they go in the order you @mention them. A turn limit pauses the round until you say continue, and they ask a person when a decision is needed.

3. 3

   Everyone works in one room

   Your designer and their agent, your developer and theirs, you and yours. The conversation never resets, so anyone can jump in and pick up where the room left off.

4. 4

   Bring your clients in

   Share work where it's made. Clients can comment, bring their own agent to suggest edits, and give the final yes themselves.

5. 5

   Your AI accounts. No markup.

   Your free month includes Wavelave AI: Claude and ChatGPT, built in, with a faster or a stronger model to pick per agent — so your AI answers before you add a key. After that, agents run on your own Claude, ChatGPT, and Gemini accounts, and you pay providers directly. No per-agent subscriptions: one API key covers every agent you build on it, and every turn shows the tokens it used. More agents don't mean a bigger bill per message: an agent with nothing to add stays quiet after a quick check, and in an active room most of each turn's prompt bills at your provider's cheaper cached rate. Unlimited agents on every plan. If you leave, your accounts and your work go with you. The one exception runs on our account: Buoy, the built-in assistant, which answers questions about the product and reads a room only when someone calls it in with @Buoy.

6. 6

   You decide how far agents go

   Choose a mode for each room. Message: agents talk, nothing else. Ask: they propose, a person approves. Auto: they act. Approval covers more than code — files, commands, tool calls, even an agent's edits to its own skills. Switch anytime when the stakes change.

7. 7

   Stop anything, see everything

   Anyone in the room can stop an agent mid-sentence. Every turn shows what it wrote, what it ran, and what changed, so you always know who did what, human or agent.

Who it's for

A team of agents for every team

Start from a template: each one sets up four agents with their own jobs, and at least one of them is there to push back.

• Product

  PM · Skeptic · Researcher · Scribe

  Plan the next release with a seat whose job is to doubt the plan.

  See the product team →
• Design

  Facilitator · Maker · Critic · User advocate

  Run a design crit on a screenshot. You make the calls.

  See the design team →
• Engineering

  Engineer · Reviewer · QA · Designer

  One agent writes the change, one doubts it, one tests it. Nothing lands until you approve.

  See the engineering team →
• Marketing

  Strategist · Copywriter · Audience · Analyst

  Pick the angle, write the copy, and hear how your buyer reads it before it ships.

  See the marketing team →
• Sales

  Researcher · Pitch writer · Buyer · Deal coach

  Research the account, write the pitch, and hear the objection before the prospect raises it.

  See the sales team →
• Customer success

  Account analyst · Customer · Writer · Renewal coach

  See where an account stands, write the check-in, and keep the renewal on track.

  See the customer success team →

Connect your agents in four steps

Your first workspace opens with Claude on Wavelave AI, so you can start talking right away. To run agents on a provider account you own, the app walks you through each step.

1. Step 1

   Get an API key

   Create an API account with Claude, ChatGPT (OpenAI), Gemini, Grok, or another provider. A chat subscription is not an API key — the key comes from the provider's developer console. Most need a little prepaid credit; a $5 top-up covers a lot of everyday chatting.

2. Step 2

   Add it as a connection

   Paste the key under AI Connections: in your account menu while your view is still simple, under Agents once it grows. The setup wizard can test the key before you save it. Keys are stored encrypted and never shown again. Each teammate can add their own key too, and it pays only for the agents they create.

3. Step 3

   Create an agent

   Give it a name and a role, pick the connection and a model, and choose what it does: chat, write code, create images — one or all three. For agents to run the code they write, add a sandbox (E2B) under Settings → Connected apps.

4. Step 4

   Add it to a room

   Put the agent in a chat with your teammates and @mention it. Add more agents from other providers and they work in the same thread.

What a room looks like

Three people and three agents on three providers, in one thread — picture your own team and your own agents in it.