Anthropic Integrates Claude Cowork Into Chrome Extension, Enabling Skills and Plugins in Browser
Anthropic's Chrome extension now runs full Claude Cowork sessions in its side panel, letting skills, plugins, and connectors operate directly in the browser. The update is live for all paid plans via the Chrome Web Store.
Anthropic has integrated Claude Cowork directly into the side panel of its Chrome extension, enabling skills, plugins, and connectors to run in the browser for the first time without additional setup.
What changed
Anthropic's Chrome extension already allowed Claude to read logged-in pages, click, type, and fill out forms — but only for browsing and information gathering. With this update, a complete Cowork session now runs inside the same side panel. According to Anthropic, this means Claude in Chrome can handle research while Cowork converts the output into deliverables like Excel spreadsheets, PowerPoint slides, or written reports.
Anthropic lists several use cases for the integration: pulling metrics from analytics dashboards, organizing files in Google Drive, and logging sales calls in Salesforce. The company says Claude will ask for user confirmation before completing any purchases during a session.
Sessions are tied to the user's Anthropic account rather than a specific device, according to the company.
Security caveats
Anthropic reiterated warnings about prompt injection risks tied to browser-based agents. The company states that hidden instructions embedded in websites remain a potential attack vector for the extension, consistent with risks seen across LLM-powered browser interfaces industry-wide. Anthropic advises against using the extension with banking or health data.
Availability
The update is live now for all paid Claude plans through the Chrome Web Store. Anthropic has not disclosed specific pricing changes tied to this feature — it is bundled into existing paid subscription tiers.
What this means
This is a packaging and integration update, not a new model release — Claude's underlying weights are unchanged. The significance lies in workflow consolidation: users no longer need to switch between a research-only browser agent and a separate Cowork session to produce structured outputs like spreadsheets or slide decks. Folding skills, plugins, and connectors into the browser context extends Anthropic's push toward agentic browsing, following moves by OpenAI and others to make browser-based agents handle multi-step tasks end-to-end.
The security caveats are worth noting closely. Prompt injection remains an unsolved problem across the industry, and running full Cowork sessions — with connector access to tools like Salesforce and Google Drive — inside a browser that can read arbitrary web pages raises the stakes if a malicious page manages to inject hidden instructions. Anthropic's explicit warning against using the tool with banking or health data signals the company still views this as an early-stage capability rather than one ready for high-sensitivity data handling.
Related Articles
Anthropic Report: Claude Was Used to Target US Navy Ships, Build Missiles, and Track Uyghurs
Anthropic's latest threat intelligence report documents five cases where state and non-state actors used Claude for military targeting, weapons development, mass surveillance, and repression. The findings include an Iran-linked operation targeting US naval forces and a Mali-based system capable of monitoring 25 million phones.
Anthropic Threat Report: Claude Used for Missile Software, Mass Surveillance, and Systematic Theft by Chinese AI Labs
Anthropic's latest threat intelligence report covers December 2025 through August 2026, documenting Claude's misuse in espionage, weapons development, and nationwide surveillance operations. The report also details how seven Chinese AI labs ran covert networks—some routing their own customers' requests through Claude—to extract training data at industrial scale.
Analysis: Claude 'Fable 5.1' Drops Em Dashes and Hedging Language, Answers Grow 30% Longer
A new Arena.ai analysis of tens of thousands of Text Arena outputs shows Claude 'Fable 5.1' has shifted its writing style significantly from Fable 5 — using fewer em dashes, less hedging language, and producing 30% longer responses. The codenamed models appear to be unreleased Anthropic checkpoints being tested anonymously on LMArena.
Anthropic CEO Dario Amodei Proposes Three-Step Plan to Deliberately Slow AI Capability Advances
Anthropic CEO Dario Amodei published an essay proposing a three-step plan to deliberately pace AI development, including third-party safety audits and cross-industry coordination. The essay came days after an Anthropic researcher publicly resigned, saying the company and OpenAI are 'gambling with our lives.'
Comments
Loading...