Claude for Chrome
Also known as: Claude in Chrome, Claude browser extension, Anthropic browser agent, Claude Code Chrome extension
Anthropic's browser agent: a Chrome side panel that navigates, clicks, and fills forms with CDP level control, records actions into reusable scheduled shortcuts, bridges to Claude Code and Desktop, and leads the category on safety posture (high risk action approvals, site permissions, published prompt injection defenses). Beta on all paid plans.
Claude for Chrome is Anthropic's browser agent: a side panel extension that lets Claude navigate, click buttons, and fill forms on real pages, pairing with Claude Code, Cowork, and Claude Desktop so browser work feeds directly into coding, documents, and analysis.
Third party analysis of the extension documents a recording to workflow system: user actions are captured (DOM snapshots and mouse events via rrweb), optionally narrated by voice, and converted into reusable parameterized shortcuts that can be scheduled for recurring daily, weekly, monthly, or annual execution, with external MCP server connections and a native messaging bridge to the local Claude Desktop or CLI.
Safety is the product's signature: permission is required for high risk actions, site level access is user managed, an ask before acting mode is available, and Anthropic launched deliberately slowly (a 1,000 user Max research preview in August 2025) while publishing prompt injection defenses that significantly reduced attack success; first party materials still label the beta a feature with unique risks requiring alertness.
It drives the browser at CDP level through the chrome.debugger API, and model access varies by plan, with higher tiers choosing among Anthropic models by task complexity. Indexed under the incumbent ruling as a named, independently installable, documented product; graded from retrieved evidence like any vendor.
Vendor details
Canonical URL
https://claude.com/claude-for-chrome
Category
Browser / computer-use agent
Funding status
Product of Anthropic; launched as a research preview to 1,000 Max subscribers in August 2025 with staged prompt injection defenses, now in beta on all paid Claude plans
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Works with Claude Code (claude --chrome), Claude Desktop, Cowork, and VS Code for end to end workflows via a native messaging bridge; connects to external MCP servers; drives Chrome and Edge through the chrome.debugger API (CDP), synthesizing trusted-level clicks and keystrokes; manages tab groups per session with agent created secondary tabs.
Sources & related URLs
Agentic Index coverage score
7.5 / 14 capabilities · 54%
| Integrations & Tool Calling | Full |
|---|---|
|
Pairs with Claude Code, Cowork, Claude Desktop, and VS Code for end to end workflows, connects to external MCP servers, and acts on any site in Chrome and Edge, Anthropic materials and extension analysis 2026-07-22 |
|
| Workflow Orchestration | Partial |
|
Multi step in browser task execution with recorded workflows converted into parameterized reusable shortcuts, short of a general workflow orchestrator, Anthropic materials and extension analysis 2026-07-22 |
|
| Knowledge Grounding & RAG | Unable to verify |
|
Session context follows browsing activity but no knowledge grounding or retrieval layer is documented, Anthropic materials 2026-07-22 |
|
| Human Oversight & Guardrails | Full |
|
Permission is required for high risk actions, site level access is user managed, an ask before acting mode is available, and skipping permissions for trusted workflows carries explicit supervision guidance, Anthropic materials 2026-07-22 |
|
| Security, Identity & Governance | Full |
|
Prompt injection defenses with measured reductions in attack success, per site permission management, a deliberately staged rollout, and candid first party risk labeling of the beta, Anthropic materials and press 2026-07-22 |
|
| Observability & Auditability | Unable to verify |
|
No run logging, replay, or audit tooling documented for users or administrators, Anthropic materials 2026-07-22 |
|
| Memory & State Persistence | Unable to verify |
|
Session tab group state is per conversation and reusable shortcuts are covered under packaging; no persistent agent memory layer documented, extension analysis 2026-07-22 |
|
| Deployment & Data Residency | Unable to verify |
|
Extension plus cloud models; the native messaging bridge to local Claude Desktop or CLI is architecture rather than a deployment or residency option, extension analysis 2026-07-22 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Recording user actions (with optional voice narration) converts into reusable parameterized shortcuts saved as a prompt library with scheduling, a first class recording to workflow packaging system, extension analysis 2026-07-22 |
|
| Triggers & Channel Coverage | Partial |
|
Saved shortcuts can be scheduled for recurring daily, weekly, monthly, or annual execution per extension internals analysis; no broader event trigger channels documented, extension analysis 2026-07-22 |
|
| Model Flexibility & Routing | Partial |
|
Higher plan tiers choose among Anthropic models by task complexity, a documented model selection surface limited to one provider family, integration guide 2026-07-22 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
External MCP server connections, a native messaging bridge to Claude Desktop and the Claude Code CLI (claude --chrome), and VS Code availability, extension analysis and integration guide 2026-07-22 |
|
| Testing, Debugging & Optimization | Unable to verify |
|
No agent testing, simulation, or optimization tooling documented, Anthropic materials 2026-07-22 |
|
| Browser & Computer Use | Full |
|
SourceThe product is browser control: navigating, clicking, and filling forms through chrome.debugger CDP access that synthesizes trusted level input events, Anthropic materials and CHEQ analysisread 2026-07-22 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Unable to verify against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Recent platform changes
Claude in Chrome is now generally available for all paid Claude plans. The extension can now take actions autonomously in the browser without requiring user approval for every step, relying on a safety classifier to validate each action before execution.
Bears on: Browser/computer use
View sourceAnthropic introduced cross-surface memory for Claude Cowork, which powers the Claude in Chrome side panel. Users can now manage, edit, or delete remembered topics, with sensitive topics excluded by default.
Bears on: Memory / state
View sourceAnthropic updated the Claude in Chrome extension to operate as a full Claude Cowork session, enabling session synchronization across desktop, web, and mobile platforms. The update automatically integrates a user's skills and connectors into the browser environment and adds a mandatory human-approval step before the agent executes high-risk actions like submitting forms.
Bears on: Memory / state
View sourcePricing
Included in paid Claude plans
included in Claude subscription plans
What is public
The bundling is public (beta on all paid Claude plans; Max preview gating at a reported $100 to $200 per month in 2025); the current entry plan rate was not captured in retrieved materials.
Variable cost rationale
Flat subscription bundling; model usage limits vary by plan tier rather than metered per action billing.
Sales call required
No, self serve available
Free / trial
Not available on free plan per retrieved materials
Related vendors
- AgentQL — A self-healing query language for the web: natural language…
- Airtop — Cloud browsers for AI agents that compile plain English workflows…
- Asteroid — Supervised browser and computer use agents for regulated back…
- Astral — Browser based AI marketing agent that operates social platforms like…
- Automat AI — Managed AI agents that operate computers visually to replace legacy…
- Autotab — General AI agent that uses a mouse and keyboard like a person to do…
Alternatives to Claude for Chrome
The closest documented capability profiles to Claude for Chrome among browser and computer-use agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Island11.0 / 14Adds documented Knowledge Grounding & RAG and Observability & Auditability, among others
- Scrapybara9.0 / 14Adds documented Observability & Auditability and Memory & State Persistence, among others
- Brave Leo AI Browsing6.5 / 14Adds documented Knowledge Grounding & RAG and Deployment & Data Residency
- Manus7.5 / 14Adds documented Knowledge Grounding & RAG and Observability & Auditability, among others
- Minded7.5 / 14Adds documented Observability & Auditability and Deployment & Data ResidencyClaude for Chrome vs Minded →
- Twin9.5 / 14Adds documented Knowledge Grounding & RAG and Observability & Auditability, among others
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded