Claude Cowork
Claude's agent for office work: point it at your folders and tools, give it a goal, and it does the steps.
By Anthropic · launched 2026-01-12 · last checked 2026-11-01
Why 4: it plans and works through multi-step tasks, using sub-agents and a browser, and can run scheduled tasks unattended; deleting anything still needs your approval, so it is not a 5.
What it does. Cowork launched on 12 January 2026 as a research preview for Max subscribers on macOS, built on the Claude Code harness for non-coding work. You choose which folders and tools it can reach; it plans, works through the steps in view, can use sub-agents and a built-in browser, and can run scheduled tasks unattended. Since 6 October 2026 new tasks on Pro, Max and Ultra run in the cloud, reaching local folders only through the desktop app while it is open.
Where it falls short. The defaults are careful ("Deleting anything needs your approval"), but computer use is still a research preview and Anthropic warns that Cowork takes real actions on your files. On the Remote Labor Index, Claude Opus 4.6 in the Cowork harness automated 4.17% of projects.
What it can do
Incidents
- 2026-01-15 Days after launch, PromptArmor showed that a hidden instruction in a document could make Cowork upload a user's files to an attacker's Anthropic account without further approval; Anthropic advised caution and said it was shipping a fix to its virtual machine. Source
- 2026-07-01 Pentera Labs showed that a prompt planted in a compromised account's synced preferences could turn Claude Desktop into a remote-control tool on a developer's machine; Anthropic said this was not a vulnerability in its product because the account was already compromised. Source
Closest fiction
Robot & Frank — Polite, follows the routine, asks before it acts, and helps whoever is talking to it.
Related general assistant agents
- ChatGPT Work OpenAI · autonomy 4/5
- Gemini Spark (Gemini Agent) Google · autonomy 4/5
- Genspark Super Agent Genspark (MainFunc) · autonomy 4/5
- Grok Bot xAI · autonomy 4/5
- Manus Butterfly Effect · autonomy 4/5
- Perplexity Computer Perplexity · autonomy 4/5
History
- 2026-11-01 Added to the index.
Sources
- Claude Cowork Anthropic
- Use Claude Cowork on web, desktop and mobile Anthropic (Claude Help Center)
- Contagious Claude Code bug Anthropic ignored promptly spreads to Cowork The Register, 2026-01-15
- Red teamers turned Claude Desktop into a double agent to do their evil bidding The Register, 2026-07-01
- Remote Labor Index leaderboard Scale AI and Center for AI Safety
← All agents in the index · Autonomy is rated on what the product does today, on a 1–5 scale. Every fact links to a source.