13 · computer-use · 06 systems
Browser & computer-use agentsAgents that drive a GUI — clicks, forms, and logins, not commits.
Computer-use agents operate software through its interface — a browser, a desktop, a remote session — by reading the screen or DOM and acting with clicks, keystrokes, and navigation. Some are hosted products that take a goal and return a result (ChatGPT Agent, Manus); others are open frameworks and APIs for building your own (Browser Use, Skyvern, Stagehand, Claude Computer Use).
Systems in this category
Browser Use
Browser Use · computer-use
Browser Use is the breakout open-source framework for browser agents: connect an LLM to a controllable browser, describe a task, and let it click, type, and navigate to completion — self-hosted or via Browser Use Cloud.
Skyvern
Skyvern · computer-use
Skyvern is an open-source browser-automation agent that uses LLMs plus computer vision to operate websites it's never seen — no brittle per-site selectors.
ChatGPT Agent
OpenAI · computer-use
ChatGPT Agent is OpenAI's hosted computer-use mode — the Operator lineage: it spins up a managed browser, works through multi-step web tasks, and can take actions after confirmation.
Claude Computer Use
Anthropic · computer-use
Claude Computer Use is Anthropic's API capability: Claude reads screenshots and emits mouse/keyboard actions, letting developers build agents that operate desktops and browsers in their own sandboxes.
Manus
Manus · computer-use
Manus is the hosted general agent: give it a goal in chat and it works a cloud session — browsing, running code, building artifacts — returning finished deliverables rather than suggestions.
Stagehand
Browserbase · computer-use
Stagehand is Browserbase's open-source agent framework: act/extract/observe primitives over a controllable browser, mixing deterministic Playwright code with LLM-driven steps.
Who it's for
Teams automating work that lives behind a UI with no API — web back-office tasks, data entry, QA flows, form-filling, scraping-with-login — and developers building agent products on top of browser control.
Tradeoffs
GUI automation is slower and flakier than APIs: sessions break on UI changes, CAPTCHAs and bot detection are constant friction, and giving an agent a logged-in session raises real credentialing and audit questions.
Frequently asked
What is a computer-use agent?
An agent that operates a graphical interface the way a person does — reading the screen or accessibility tree and emitting clicks and keystrokes — rather than calling an API. The term comes from Anthropic's Computer Use capability; OpenAI's Operator (now ChatGPT agent mode) popularized the hosted version.
Who should use a computer-use agent?
Teams automating work that lives behind a UI with no API — web back-office tasks, data entry, QA flows, form-filling, scraping-with-login — and developers building agent products on top of browser control.
What are the tradeoffs of computer-use agents?
GUI automation is slower and flakier than APIs: sessions break on UI changes, CAPTCHAs and bot detection are constant friction, and giving an agent a logged-in session raises real credentialing and audit questions.
Which computer-use agents are in the catalog?
6 as of 2026-09-25: Browser Use (Browser Use), Skyvern (Skyvern), ChatGPT Agent (OpenAI), Claude Computer Use (Anthropic), Manus (Manus), Stagehand (Browserbase).
Hosted agent vs framework — which do I need?
Hosted agents (ChatGPT Agent, Manus) take a goal in a chat box and work in a managed browser — good for ad-hoc tasks. Frameworks (Browser Use, Skyvern, Stagehand) are code you run against your own browser sessions — good when the automation is a product feature or runs on a schedule.
How do computer-use agents differ from coding agents?
Coding agents act on a repository through an editor or shell. Computer-use agents act on any application through its GUI — the target is software that was never designed for automation, which is why the category exists alongside coding agents rather than inside them.