Stop playing telephone between Claude and Codex

What it does

Tutti is a local-first, GUI-based workspace that sits next to your existing AI agent subscriptions. It gives Claude Code, Codex, and similar tools a shared environment where they can see each other’s conversations, files, and running tasks. Instead of copying API docs or design assets from one chat window to another, you reference prior outputs with @ and + inside a single interface. Everything stays on your machine; the open-source version keeps working state local.

The interesting bit

The project treats agent handoffs as the boring part worth solving. Tutti bundles its own app ecosystem—image generation, UI prototyping, slide decks, docs—that both you and your agents can invoke. When Codex needs a mockup, it can call the design app directly and drop the result into its workspace without you downloading, uploading, or re-explaining requirements. It is essentially glue code, but it’s glue code that replaces an entire class of human copy-paste labor.

Key highlights

• @ references let one agent pull in another’s past conversations, files, or app invocations without manual context dumps.
• Built-in app center (AI Canvas, AI PPT, prototype design, docs) that agents can invoke as part of a workflow.
• Goal-to-task decomposition: describe a high-level goal and Tutti breaks it into sub-tasks assignable to different agents.
• Single Control Center view for pending approvals, running tasks, and active agent conversations.
• Reuses your existing Claude Code and Codex subscriptions; it does not resell model access.

Caveats

• The open-source build is explicitly single-user; multi-user and multi-device collaboration are reserved for the upcoming Tutti · VM cloud offering.
• Only Claude Code and Codex are currently supported; Gemini, OpenClaw, and Hermes integrations are listed as coming soon.
• Several features and the VM variant are on a waitlist or in development, so the current open-source scope is narrower than the full vision.

Verdict

Worth a look if you’re already juggling multiple agent subscriptions and losing time to context handoffs. Skip it if you only use one agent, prefer terminal-only workflows, or need true multi-user collaboration today.