Product · Your AI agent

Your agent. Your subscription. Your machine.

Bring Claude Code, Codex, Copilot or OpenCode. It runs on your computer with your own login, so agent work runs on the plan you already pay for. IdeaWave never sees your credentials.

Works with

Works with the agent you already use

Pick the runtime and model for each agent card. The model list comes from your own subscription.

  • Claude Code

    Sign in from the runtime menu, or with claude login or an API key.

  • Codex

    Sign in from the runtime menu, or with codex login or an API key.

  • Copilot

    GitHub Copilot CLI. Sign in with copilot login on a Copilot plan.

  • OpenCode

    Sign in with opencode auth login for the provider you already use.

  • Any ACP agent

    Add a custom Agent Client Protocol runtime from the runtime menu.

Two ways to connect

Both run the same small host on your machine. It listens only on your computer's loopback address and passes work between the board and your agent.

The desktop app

IdeaWave with the local host built in. Open it, sign in, and your agent is connected. No terminal, and it keeps itself up to date.

  • macOS (Apple Silicon) and Linux x64
  • Windows support is coming soon
  • Best for browser cards

One command

Keep using your own browser and run the local host from a terminal. It connects to the board you have open.

  • macOS, Linux and Windows x64
  • Needs Node.js 22 or newer

Reads your code. Never edits it.

Attach a folder and your agent plans from the real code, not a guess from an old doc. It can read. It can't write files, edit them or run commands, whatever agent you use.
  • You choose which folders it can read
  • Writes, edits, shell and terminal are rejected for every runtime
  • Plans cite the actual files they're based on

Your login. Your plan.

Agent cards run on the account or keys you already use with Claude Code, Codex, Copilot or OpenCode. IdeaWave doesn't resell AI, so there's no second bill for agent work.
  • Sign in to Claude Code or Codex right from the runtime menu
  • Model lists refresh from your own subscription
  • Pick the runtime, model and folder per agent card

Hand the plan to your coding agent

IdeaWave plans. When the plan is agreed, Claude Code or any MCP client reads the board from your terminal and builds it in your repo, with no tab open.
  • A personal token gives read-only access to your boards
  • Revoke it any time in Settings

What's next

Labelled honestly. Neither is available yet.

  • Beta · coming soon

    Background agents

    An agent that watches the board and suggests the next question, flowchart or bit of research. Nothing lands until you approve it.

  • Beta · coming soon

    Voice agent

    Talk to the board and have your agent arrange, research and build while you think out loud.

Questions about your agent

Who needs what, and what goes where.

Does everyone on the team need an agent?

No. One person with an agent connected is enough to get going. Everyone else opens the board in a browser, answers questions and edits the plan.

Does IdeaWave see my AI login or API keys?

No. Your agent runs on your machine and signs in the way it always does. IdeaWave never receives those credentials.

Where does what the agent reads go?

To your model provider, under your own plan, exactly as when you use the agent anywhere else. The board itself is hosted by us so your team sees it live. Local files are only shared when you add them to the board.

Can it change my code?

No. It can read the folders you attach, but writing files, editing and running commands are switched off for every agent. To build the plan, hand it to your coding agent in your own terminal.

Do you charge for agent usage?

No. Agent work runs on the subscription or keys you already have. The one exception is generating images on the board: each team gets 10 free images a month, then they use credits.

Bring the agent you already pay for.

Open a board, then connect with the desktop app or one command.