Our verdict · Coding agent
The most capable agent here for work that spans a real codebase: it plans, edits many files, runs tests, commits and opens pull requests, from the terminal, an IDE, the desktop app or a cloud session. It comes with every paid Claude plan, so cost is flat, but usage limits reset every five hours and weekly. It has no hosting or visual editor, so it's slower than a builder to a first demo.
Pick it if
- Developers and teams with an existing repository
- Multi-file changes, refactors, tests and CI automation
- People who want one agent across terminal, IDE, desktop, web and phone
Skip it if
- You want a hosted app without touching code or a terminal
- You need models from other vendors in the same tool
- Your heavy days won't fit in five-hour usage windows and you won't pay for Max
Our history with it: This site was built with Claude Code, and Mahmoud uses it regularly. Anthropic has never paid us and we take no commission.
Claude Code is Anthropic's agentic coding tool. It reads a codebase, plans, edits many files, runs commands and tests, commits and opens pull requests, and it does that from the terminal, VS Code, JetBrains, a desktop app, the web and the Claude mobile app, Slack, and CI. This site was built with it and Mahmoud uses it regularly, so read the strengths with that in mind; prices and limits come from Anthropic's pricing page and the Claude Code docs, linked below.
Scores
- Speed to a first version3/5
Fast once a repo exists; a first visual demo is slower than a builder because there's no hosting or canvas.
- Depth before the wall5/5
Subagents, skills, hooks, MCP, plugins and scheduled cloud tasks, in large real codebases.
- Backend and data2/5
Brings no backend; it sets up and works on whatever you choose.
- Ownership and exit5/5
Everything stays in your repo and infrastructure; CLAUDE.md can be AGENTS.md shared with other agents.
- Pricing predictability3/5
Flat, published plan prices; the limits are multiples of Pro rather than published numbers.
- Team and production readiness4/5
Team and Enterprise seats, GitHub code review, CI integrations; Enterprise usage billed at API rates.
Overall 3.7/5, the plain average of six criteria. Scored by Mahmoud Halat from published specs and docs, checked Nov 1, 2026. How we score.
Pricing and the usage windows
| Plan | Price (USD) | What you get |
|---|---|---|
| Pro | $20/mo ($17/mo annual) | Claude Code included, baseline usage |
| Max | $100/mo | 5x Pro usage every 5 hours and every week |
| Ultra | $200/mo | 10x Pro weekly, no 5-hour limit for verified accounts |
| Team Standard | $25/seat/mo ($20 annual) | Claude Code included |
| Team Premium | $125/seat/mo ($100 annual) | 5x Standard usage |
| Enterprise | $20/seat/mo annual + API usage | 20-seat minimum; usage billed at API rates |
- Billed in
- Flat subscription with usage limits that reset every 5 hours and weekly
- Free tier
- Not included on the Free plan
Source: claude.com/pricing. Prices checked Nov 1, 2026, before tax. It can also run on an Anthropic Console API key, billed per token, or through third-party providers in the CLI and IDE extensions.
Claude Code is priced as part of a Claude subscription, which makes it the easiest agent here to budget: the monthly price is fixed. What varies is how much you can do before you hit a limit. Usage resets every five hours and every week, and Anthropic describes the plans as multiples of Pro rather than in tokens:
- Pro ($20, or $17 billed annually) covers regular work on one project in our experience, with occasional waits on heavy days.
- Max ($100) gives 5x Pro in both windows and suits daily, multi-hour agent sessions.
- Ultra ($200) gives 10x Pro weekly and removes the five-hour window for verified accounts.
- Team Premium seats ($125 monthly, $100 annual) give 5x Standard, for engineers who run agents all day.
- Enterprise ($20 a seat a month billed annually, 20-seat minimum) bills usage at API rates on top, which trades the flat price for predictability by usage.
For automation that runs unattended, an Anthropic Console API key billed per token is usually the cleaner option than a personal subscription.
What it does better than an editor agent
Claude Code's advantage is that it's an agent first and an interface second. It works across a whole repository, runs your tests and fixes what fails, and can split a job across subagents that work in parallel while a lead agent merges the results. It's extensible in the same open formats as the rest of this site's ladder: CLAUDE.md for project instructions (or an existing AGENTS.md), skills for repeatable workflows, hooks to run your formatter or linter, MCP servers for outside systems, and plugins that bundle them.
The surfaces are what make it practical for a team. Start a task in the terminal, continue it on the desktop app with visual diffs, kick off long tasks on the web and check them from a phone, route a bug report from Slack into a pull request, run reviews in GitHub Actions or GitLab CI, and schedule recurring jobs that run in the cloud while your computer is off. If you outgrow all of that, the Agent SDK exposes the same loop as a library.
Where it struggles
Claude Code has no hosting and no visual canvas, so it's slower than Lovable or Bolt to a first demo, and it assumes someone can review what it changes. It runs Claude models only. And the limits are the price of the flat fee: a long refactor on Pro can stop for a few hours mid-task, and the limits are published as multiples, not numbers. Teams that need guaranteed throughput end up on Max, Team Premium or API billing.
Ownership
Everything happens in your repository, on your machine or in a cloud session attached to a repo you choose. There's nothing to export, and the project instructions, skills and MCP configuration are plain files you can reuse with other agents.
Key facts
- Made by
- Anthropic
- Works in
- Terminal, VS Code, JetBrains, desktop app, web and mobile, Slack, GitHub Actions and GitLab CI
- Models
- Claude models only
- Backend
- None: it works on whatever stack your repo uses
- Ownership
- Your repository, your machine or a cloud session on a repo you choose
Alternatives
- CursorIf you'd rather stay in an editor and review each diff
- OpenAI CodexThe same shape from OpenAI, with published per-model limits
- GitHub CopilotCheaper seat for issue-to-PR work inside GitHub
- Windsurf (Devin Desktop)An editor that can host Claude Agent and other agents side by side
Head-to-head: Cursor vs Claude Code · Claude Code vs Codex
Questions people ask
Is Claude Code free?
No. Claude Code isn't included on Claude's Free plan. It comes with every paid plan, from Pro at $20 a month ($17 billed annually) through Max at $100, Ultra at $200 and Team seats. It can also run on a pay-per-token Anthropic Console API key. Checked November 1, 2026.
How do Claude Code's usage limits work?
Subscription usage resets every five hours and every week. Max gives 5 times Pro's usage in both windows, and Ultra gives 10 times Pro weekly with no five-hour limit for verified accounts. Anthropic publishes these as multiples of Pro rather than as token or message counts.
Does Claude Code have a GUI?
Yes, several. Besides the terminal CLI there are VS Code and JetBrains extensions, a desktop app with visual diffs and parallel sessions, and Claude Code on the web and in the Claude mobile apps. All of them use the same engine and read the same CLAUDE.md, settings and MCP servers.
Can Claude Code use GPT or Gemini models?
No. It runs Claude models, through Anthropic or through third-party providers such as Amazon Bedrock and Google Vertex AI in the CLI and IDE extensions. If you want several vendors' models in one tool, Cursor or Devin Desktop fit better.