The archive · Product Ideas · Product decision · 2025
Anthropic Claude Code (2025): an AI engineer that lives in your terminal
Claude Code puts a coding agent in the terminal: Claude reads the repo, edits files, runs tests and commits, keeping the human in the loop.
Anthropic
What it had to solve
AI coding tools had been autocomplete-style assistants that finished lines as developers typed. Anthropic's own engineers found Claude could carry out substantial engineering tasks, so the company packaged that internal capability as a command-line collaborator for real repositories.
How it works
On February 24, 2025, Anthropic announced Claude Code alongside Claude 3.7 Sonnet: a command-line tool for agentic coding, available first as a limited research preview. Instead of completing lines as a developer typed, Claude Code took delegated engineering tasks and worked through them, from reading the codebase to committing the result.
The tool is an active collaborator: it can search and read code, edit files, write and run tests, commit and push to GitHub, and use command-line tools — keeping the developer in the loop at every step. In Anthropic's early testing, Claude Code completed in a single pass tasks that would normally take more than 45 minutes of manual work, and it became indispensable inside the company for test-driven development, debugging and large-scale refactoring.
The research preview doubled as a design loop: developers who joined got the same tools Anthropic used internally, and their feedback directly shaped the product. On May 21, 2025, with the Claude 4 launch, Claude Code became generally available, adding native VS Code and JetBrains extensions, background tasks via GitHub Actions, and an extensible SDK.
The bet paid off quickly. By October 2025, TechCrunch reported Claude Code had grown 10x in users since May and generated more than $500 million in annualized revenue for Anthropic, and the company said about 90% of the product's own code is written by Claude. The release marked the industry's shift from autocomplete-style assistants to agents that manage engineering work.
Why it lands
- The terminal is where developers already work, so adoption needed no new surface or habit.
- It delegated whole tasks — tests, debugging, refactors — instead of suggesting lines, changing the developer's job from typing to reviewing.
- It shipped from internal need: Anthropic built the tool it depended on, so design decisions came from real engineering pressure rather than speculation.
- The research-preview launch turned early adopters into co-designers, letting Anthropic tune the agent on live feedback before general availability.
What it did
By October 2025, Claude Code had grown 10x in users since its May launch and accounted for more than $500 million of Anthropic's revenue on an annualized basis; Anthropic says 90% of the product is written by its own AI models.
What you can take
Your most useful internal tool can become your best product: the team shipped the thing it depended on daily, and real usage — not roadmap debate — decided what it became.
Since then
Claude Code became Anthropic's flagship product and a centerpiece of the company's growth: Anthropic extended it to the web and the Claude iOS app in October 2025 so developers could run agents from a browser, and reports through late 2025 put the tool near $1 billion in annualized revenue. It also reset the category — GitHub Copilot, Google and OpenAI all shipped or accelerated agentic coding tools in its wake, and Anthropic credited Claude Code with helping it overtake rivals in enterprise AI adoption.
Sources
spotted an error? The archive wants to know.
Your turn
You just read one. Describe the brief you are staring at, and see who has been given the same problem.
Free account · 3 free questions · no card