Anthropic used its May 6 Code with Claude event to show a long list of enterprise moves, from SpaceX compute to Wall Street agent templates and Microsoft 365 integrations. The quieter signal was Orbit, a new assistant for Claude Cowork.
The basic idea is simple: Claude should not wait for a prompt. By the time a user starts the day, Orbit can already have prepared a briefing on what needs attention, what changed overnight and what should be answered.
What Orbit Is Built To Do
Orbit sits inside Claude Cowork as a proactive assistant. Unlike a conventional chatbot that responds only after a window is opened, it is designed to run in the background.
- It connects to Gmail, Slack, GitHub, Calendar, Drive and Figma.
- It is aware of the user's time zone.
- It can generate a personalized briefing in the morning or at a scheduled time.
- The briefing can cover emails that need replies, Slack discussions the user missed, GitHub changes from the previous night, Figma updates and meetings for the day.
"Users will get personalized insights from Gmail, Slack, GitHub, Calendar, Drive, Figma, and other apps, which Claude will generate proactively."
The GitHub and Figma connectors matter. This is not just an executive inbox-and-calendar briefing in the style of OpenAI Pulse. Orbit is aimed squarely at people who build products: developers, designers and product managers.
Why This Is More Than Another Briefing Feature
Over the past half year, the major AI labs have been moving toward the same product shape.
| Company | Product | Timing | Primary audience |
|---|---|---|---|
| OpenAI | ChatGPT Pulse | September 2025 | Executives and administrative users |
| Remy / Gemini Agent | Around I/O 2026 | General users and a 24/7 digital companion | |
| Anthropic | Orbit | May 2026 | Developers, designers and PMs |
All three are treating proactive assistants as a likely foundation for the next generation of AI products. That is unlikely to be coincidence.
For the past two years, the AI story has largely been about answering more accurately. Benchmarks carried the narrative. But as model gaps narrow, many users cannot feel the difference between an answer rate moving from 92 percent to 94 percent.
Opening a laptop and finding that Claude has already summarized what was missed overnight feels different. The next winning product may not be the one that answers best after being asked, but the one that has already looked ahead.
Four Details Worth Watching
First, Orbit is opt-in. It does not run in the background unless the user turns it on. That mirrors the privacy posture around ChatGPT Pulse and Gemini Remy. A proactive assistant cannot be enabled by default without creating a backlash.
Second, the feature flag is called tibro enabled. Read backward, that is orbit. TestingCatalog found the flag on May 4, which suggests internal testing had already been running for at least about two weeks.
Third, the current interface points beyond a single assistant. The empty settings panel says, "Your deployed Orbit apps. Pin favorites for quick access." That implies Orbit may also support installable apps, possibly an early shape of an agent marketplace.
Fourth, analysts expect Max subscribers to get early access. Pro users are likely to wait longer.
What The Move Means
Anthropic has been accelerating in enterprise accounts over the past year, including Wall Street agents, a 30,000-employee NEC rollout, Atlassian and Snyk. But for most corporate users, Claude still behaves like a tool that is opened, queried and closed.
Orbit tries to move Claude from a tool that appears when summoned into a standing colleague embedded in daily work.
The strategy is measured, but it carries a real risk. Proactive assistants have little tolerance for model errors. If Claude answers a question badly, the user may complain and move on. If it pushes a briefing every day with wrong facts or bad prioritization, the user may simply switch it off.
That makes Orbit a test not only of product design, but of whether Claude is reliable enough to hold this role. The answer should be clearer in six months.
Sources: Anthropic is working on Orbit, its upcoming proactive assistant (TestingCatalog); Anthropic's new proactive AI assistant, Orbit, will debut at the Code with Claude conference (KuCoin / Phemex); CocoLoop; Live blog: Code w/ Claude 2026 (Simon Willison)