Claude in Chrome puts Anthropic's agent inside the browser most companies already run. It left beta in August 2026, and the release changed its safety model: a classifier now approves actions that users used to approve one by one.
Key Takeaways
- An agent in your existing Chrome. — Claude in Chrome is an extension that reads pages, clicks, types and fills forms from a side panel. It became generally available on August 26, 2026 for paid Claude plans.
- Classifier instead of clicks. — Since general availability, a safety classifier checks each action against your request and probes scan pages for injected instructions, replacing per-action approval.
- Strong admin story on Team and Enterprise. — Owners can switch it off org-wide and set site allow-lists and block-lists, which also govern the agent-owned browser in Claude Cowork.
What Claude in Chrome does
The extension opens a side panel where you describe a task. Claude reads the page, clicks, types, fills forms and moves between sites until the task is done. It works inside your normal Chrome session, with your logins, which is what makes it useful and what makes the safety design matter.
At general availability on August 26, 2026, Anthropic dropped per-action approvals. Probes now scan page content for injected instructions before Claude reads it, and a safety classifier checks each planned action against what you asked (Gigazine). The same release linked the extension to Claude Cowork, whose built-in browser is owned by the agent and does not use your own logins.
Scorecard: 15/20
| Criterion | Score | Why |
|---|---|---|
| Agent | 4/5 | Acts on sites and fills forms across multi-step tasks; desktop Chrome only. |
| Control | 4/5 | Org toggle, site allow-lists and block-lists on Team and Enterprise. |
| Safety | 4/5 | Page probes plus a per-action safety classifier; fewer human checkpoints than Gemini. |
| Reach | 3/5 | Needs a paid Claude plan and desktop Chrome. |
Pricing and plans
Claude in Chrome is included with paid Claude plans; there is no separate price. Admin controls are limited to Team and Enterprise plans (Anthropic): an org-wide toggle, plus site allow-lists and block-lists applied on top of Anthropic's default blocked categories. On Team plans the extension is on by default; HIPAA-enabled Enterprise organizations keep it off until an Owner switches it on.
Who it fits, and who should look elsewhere
- Fits: teams already on Claude Team or Enterprise that want browser automation under the same admin console, and anyone who wants the option of an agent-owned browser that never touches their real sessions.
- Look elsewhere: if you need mobile coverage (Comet runs on iOS and Android), or if policy demands a human click before every purchase or post (Gemini in Chrome keeps those with the user).
Claude in Chrome alternatives
- Perplexity Comet: a standalone agent browser on every platform, with domain-limited actions in Comet Enterprise.
- Gemini in Chrome: Google's built-in agent, with buy and post buttons left to the user.
- Opera Neon and Dia: see the full ranking.
Related questions
- Which AI browsers are worth shortlisting in 2026?
- Comet, Gemini in Chrome or Claude in Chrome: which fits?
- Should a company block, sandbox or pilot agentic browsers? (partner site CTAIO)
- How does prompt injection work? (partner site CTAIO)
Sources
- Gigazine, Claude in Chrome generally available (Aug 27, 2026).
- Anthropic, Claude in Chrome admin controls.
- Forcepoint, Gartner recommends blocking AI browsers (Dec 2025).
Is Claude in Chrome free?
No. Claude in Chrome needs a paid Claude plan. It is included with those plans rather than sold separately.
Is Claude in Chrome safe?
It is designed with more injection defenses than most browser agents: probes scan pages before Claude reads them and a safety classifier checks each action against your request. It still acts with your logged-in sessions, so set site allow-lists on Team and Enterprise plans, or use the separate Cowork browser, which does not use your logins.
Can admins control Claude in Chrome?
Yes, on Team and Enterprise plans. Owners and Primary Owners can turn it on or off for the whole organization and set allow-lists and block-lists of sites. On Team it is on by default; on HIPAA-enabled Enterprise organizations it stays off until an Owner enables it.
Ready to Find the Right AI Tools?
Browse our data-driven rankings to find the best AI tools for your team.