Anthropic brings Claude Code to the web to manage AI coding agents
Anthropic has rolled out a browser-based Claude Code in research preview, letting Pro and Max users spin up agentic coding tasks from claude.ai and hand them off to the CLI—an extension of the tool’s terminal-first roots.
Anthropic has rolled out a browser-based version of its AI coding tool, Claude Code, allowing developers to create and manage autonomous coding agents directly from claude.ai.
The web experience is available to Pro subscribers and to Max plan users, extending the product beyond its original terminal-first design.
Developers can connect GitHub, set a default environment, and submit a task; Claude then analyses the repository, makes changes, runs tests, and prepares a branch for review. Sessions can be moved seamlessly from the web to the CLI via an “Open in CLI” handoff.
The web launch has complemented Claude Code’s origins as a command-line tool that plugs into developers’ workflows and IDEs (VS Code and JetBrains), with guarded file changes and configurable behaviour.
Features and access
- Start and monitor agentic coding tasks from the Claude app on the web; optionally continue locally in the terminal.
- GitHub integration via an Anthropic proxy with scoped credentials; limited-by-default outbound network access that can be configured.
- Prebuilt environments with common language toolchains; hooks for dependency setup and environment configuration.
The web release is part of making Claude Code available “where developers work,” evolving it beyond the terminal while keeping the CLI as a core experience. In an interview cited by TechCrunch, the team said the terminal will remain the “home base” for coding agents.
The company has said usage has grown markedly since a broader release in May and that Claude Code now contributes a significant share of annualised revenue; Anthropic also claimed that most of the Claude Code product has been authored with its own AI models.
Why this matters
Putting Claude Code on the web has aligned Anthropic with a wider shift toward agentic coding assistants that can operate in cloud sandboxes, connect to source control, and produce reviewable changes.
The move has arrived amid rapid competition from OpenAI, Google and Microsoft in browser-accessible coding assistants. For instance,
OpenAI has upgraded its Codex agent with a GPT‑5‑based model, broadening access across ChatGPT tiers.
Google has updated Gemini Code Assist throughout 2025 and, most recently, has transitioned tools toward an agent mode, with ongoing enhancements to coding features.