Verdict
Claude Code suits developers and engineering teams who want one agent across terminal, IDE, desktop, web and Slack, with fine-grained permission controls and the top score on Terminal-Bench 4.0 5,7,20. The catch is unpredictable usage: plan limits aren't published as numbers, and Anthropic puts typical enterprise spend at $150 to $250 per developer per month 2,12.
Good fit if you are
- Developers who work in the terminal and want an agent that runs commands, tests and git for them
- Teams that want the same agent locally, in cloud sessions, in GitHub and in Slack
- Enterprises that need managed settings, audit logs and a zero-data-retention option
Look elsewhere if you are
- Teams that want to switch between OpenAI, Google and Anthropic models in one tool
- Buyers who need a fixed, published number of tasks per month
- Developers who only want a free tier, since the Free plan excludes Claude Code
Pros and cons
What we like
- Runs in the terminal, VS Code, JetBrains, a desktop app, the browser and the Claude mobile app, and sessions move between them 4,5.
- Six permission modes, from approve-every-action to a classifier-reviewed auto mode, plus an OS-level sandbox for shell commands 7,8.
- Checkpoints let you rewind code, conversation or both to any earlier prompt 9.
- Cloud sessions run in isolated VMs, push branches and can auto-fix pull requests after CI failures and review comments 6.
- Ranks first on the independent Terminal-Bench 4.0 leaderboard at 64.85% with Opus 5.5 20.
What to watch
- Plan limits are relative multiples with weekly caps on top, so you can't tell in advance how much work a plan buys 2,3.
- Only Claude models are supported, served by Anthropic, Amazon Bedrock, Google Cloud or Microsoft Foundry 13.
- Zero data retention is limited to qualified Enterprise accounts, needs Anthropic to switch it on, and disables cloud sessions 16.
- The Bash sandbox is off by default and doesn't run on native Windows 8.
- An npm packaging error exposed Claude Code's source code in March 2026, though Anthropic said no customer data leaked 26.
Key facts
| Made by | Anthropic |
|---|---|
| Pricing | From $20/month (Pro); Max $100-$200; Enterprise $20/seat plus API usage |
| Cost per developer | $100 a month (Max 5x, billed monthly; Pro at $20 suits lighter use) |
| Also requires | Needs a paid Claude plan or a Claude Console (API) account 2 |
| Where it runs | Terminal, IDE, Cloud (background), Web app, GitHub, Slack |
| AI models | Claude models only: Fable, Opus, Sonnet and Haiku, through Anthropic, Amazon Bedrock, Google Cloud or Microsoft Foundry 13 |
| Works with | VS Code, Cursor, JetBrains, GitHub, GitLab, Slack, Chrome, Amazon Bedrock, Google Cloud, Microsoft Foundry |
| Security reports | SOC 2 Type II, ISO 27001, ISO 42001 |
| Free trial | No free tier for Claude Code |
What Claude Code is
Claude Code is Anthropic's agentic coding tool. It reads a codebase, edits files, runs commands and works with git from the terminal, VS Code, JetBrains, a desktop app, the browser and the Claude mobile app 4,5. Every surface runs the same engine, so CLAUDE.md files, settings and MCP servers carry across 4.
It comes with every paid Claude plan and shares that plan's usage pool with chat 2. Teams can also run it on pay-as-you-go API billing or through Amazon Bedrock, Google Cloud and Microsoft Foundry 10,13. Cloud sessions, routines, Slack delegation and automatic GitHub code review take it well beyond a local CLI 5,6.
What it can do
You start a session in a project folder and describe the task. Claude Code plans, edits across files, runs tests and commands, and can commit and open pull requests 4. Permission modes set how much it does without asking: Manual asks before edits and commands, acceptEdits approves file changes, plan mode explores without editing, and auto mode lets a classifier model review actions instead of you 7. Auto mode is now the default for interactive terminal and VS Code sessions 7. An optional OS-level sandbox limits which files and network hosts shell commands can reach 8, and checkpoints let you rewind code or conversation to any earlier prompt 9.
For longer work, cloud sessions run in Anthropic-managed VMs, clone from GitHub and keep going after you disconnect; claude --teleport pulls a session back into your terminal 6. Auto-fix watches a pull request and pushes fixes for CI failures and review comments 6. Subagents, background agents and routines handle parallel and scheduled work, and MCP connects tools such as Jira, Slack and Google Drive 4. Admins push managed settings that override local configuration 10.
- Actions. Claude Code plans an approach, edits across files, runs commands and tests, and creates commits, branches and pull requests 4. Subagents, background agents and cloud routines run work in parallel or on a schedule, and MCP connects tools such as Jira, Slack and Google Drive 4.
- Setup. A one-line installer, Homebrew or WinGet sets up the CLI, and you sign in with a Claude or Console account 4. Cloud sessions reach your repositories through the Claude GitHub App or the /web-setup command 6.
- Reviewing its changes. In Manual mode you approve each edit and command, the VS Code and desktop apps show diffs for review, and cloud sessions hand work back as branches and pull requests 4,6,14.
- Analytics. Team and Enterprise admins get a dashboard of active users, accepted lines and pull requests shipped with Claude Code, with CSV export, plus OpenTelemetry for per-user token and cost data 11.
Pricing
Claude Code has no separate price. It comes with Pro at $20 a month ($17 billed annually), Max 5x at $100 and Max 20x at $200, and with Team seats at $25 standard or $125 premium per month ($20 and $100 billed annually) 2,3. Enterprise costs $20 per seat per month on annual billing plus usage at API rates 2. Limits reset every five hours with weekly caps on top, and chat draws from the same pool 2. Anthropic doesn't publish limits as numbers: Max simply gives 5x or 20x Pro's per-session usage 2. Past the limit you can turn on usage credits at API rates, such as $4 per million input and $20 per million output tokens for Opus 5.5 2. Anthropic puts enterprise API spend at about $150 to $250 per developer per month 12. For all-day use, budget for Max 5x at $100 a month.
| Plan | Price | Includes |
|---|---|---|
| Pro | $20/month ($17/month billed annually)2 | Claude Code on all surfaces; at least 5x Free usage per 5-hour session, weekly limits on top, shared with chat |
| Max 5x | $100/month2,3 | 5x Pro usage per 5-hour session, higher output limits, priority access at high traffic times |
| Max 20x | $200/month2,3 | 20x Pro usage per 5-hour session, weekly limit across all models |
| Team Standard seat | $25/seat/month ($20 billed annually)2 | Claude Code, SSO, central billing, usage analytics, admin controls; teams of 2 to 150 |
| Team Premium seat | $125/seat/month ($100 billed annually)2 | 5x the usage of a standard seat |
| Enterprise | $20/seat/month (annual) plus usage at API rates2 | SCIM, audit logs, Compliance API, custom data retention, user and org spend limits, IP allowlisting |
| Usage credits and API | Opus 5.5 $4/$20, Sonnet 5.5 $2/$10 per million input/output tokens2 | Extra usage on paid plans, or pay-as-you-go through a Claude Console account |
Security and data
Anthropic holds SOC 2 Type II, ISO 27001:2022 and ISO/IEC 42001:2023 certifications and offers a HIPAA-ready configuration with a BAA 17. Under commercial terms (Team, Enterprise, API and cloud platforms), Anthropic doesn't train models on Claude Code prompts or code 15. On Free, Pro and Max, training is on unless the user opts out, and data from users who allow training is kept for five years 2,15. Commercial data is kept for 30 days by default 15. Zero data retention is available to qualified Enterprise accounts, but Anthropic must enable it for each organization, and it turns off cloud sessions 16. Cloud sessions run in isolated VMs, GitHub credentials never enter the VM, and outbound traffic passes through a security proxy 15.
- Zero data retention. Zero data retention for qualified Claude for Enterprise accounts, enabled by Anthropic per organization; HIPAA BAA coverage through the HIPAA-ready configuration 16,17
- Data residency. US-only inference on the API at 1.1x token prices 2
- Training on your data. No training on Team, Enterprise or API data by default; Free, Pro and Max data is used for training unless the user opts out 2,15
- AI models. Anthropic, directly or through Amazon Bedrock, Google Cloud and Microsoft Foundry 13
What customers say
Independent reviewers rate Claude Code highly. It scores 4.7 from about 105 G2 reviews 21 and 4.7 from 104 Gartner Peer Insights ratings in the enterprise AI coding agents market 22, and TrustRadius users give it 9.4 out of 10 from 17 reviews 24. Reviewers praise its grasp of large codebases and flag token use on long sessions 21. On the independent Terminal-Bench 4.0 leaderboard, Claude Code with Opus 5.5 ranks first at 64.85% 20. Anthropic publishes Claude Code stories from Ramp and Pictet and quotes engineers at Intercom and Notion 1,18,19.
| Platform | Rating | Reviews | Seen |
|---|---|---|---|
| G2 (Claude Code)21 | 4.7 / 5 | 105 | 8 October 2026 |
| Gartner Peer Insights (Claude Code (Enterprise AI Coding Agents))22 | 4.7 / 5 | 104 | 8 October 2026 |
| Capterra (Claude Code)23 | 5.0 / 5 | 5 | 8 October 2026 |
Benchmarks
| Benchmark | Result | Source |
|---|---|---|
| Terminal-Bench 4.020 | 64.85% with Opus 5.5 (rank 1 of 35) | Independent leaderboard |
| Terminal-Bench 4.020 | 61.82% with Sonnet 5.5 (rank 2 of 35) | Independent leaderboard |
What Anthropic claims (not independently verified):
- Ramp: more than 1 million lines of AI-suggested code implemented in 30 days, with 50% weekly active usage across engineering 18
- Ramp: up to 80% less time spent investigating incidents 18
- Pictet: a compliance gap analysis finished in hours instead of a projected two weeks 19
- Enterprise deployments average about $13 per developer per active day and $150-$250 per developer per month 12
Concerns raised by others:
- Claude Code 2.1.88 shipped to npm with a source map that exposed about 500,000 lines of source code; Anthropic blamed a packaging error and said no customer data was exposed 26.
- Around the time of the leak, paid users reported quotas running out much faster than expected, which Anthropic treated as a high-priority investigation 26.
- Anthropic added weekly rate limits in August 2025 after some users ran Claude Code around the clock or shared accounts 25.
- G2 reviewers flag high token consumption on long sessions 21.
How we scored it
| Criterion | Why | Score |
|---|---|---|
| What it can do 30% of the total | It plans, edits across files, runs tests and opens pull requests both locally and in cloud VMs, and extends through MCP, subagents and routines 4,6. | 5/5 |
| Pricing 20% of the total | Every plan price is public, including Enterprise at $20 a seat plus API rates, but usage limits are relative multiples rather than numbers 2,3. | 4/5 |
| Control 20% of the total | It has permission modes with allow and deny rules 7, an OS-level sandbox 8, diff review and pull-request delivery 4,6, checkpoints 9, and managed settings with Enterprise audit logs 2,10. | 5/5 |
| Security and data 15% of the total | Anthropic holds SOC 2 Type II, ISO 27001 and ISO 42001 17, offers zero data retention to qualified Enterprise accounts 16, and doesn't train on commercial data by default 15. | 5/5 |
| Customer evidence 15% of the total | Claude Code ranks first on the independent Terminal-Bench 4.0 leaderboard 20, scores 4.7 on G2 and Gartner Peer Insights 21,22, and Ramp and Pictet publish named results 18,19. | 5/5 |
| Total | Weighted sum of the 5 levels, out of 10. | 9.6 |
Sources
- Claude Code
- Plans & Pricing
- What is the Max plan?
- Overview
- Platforms and integrations
- Use Claude Code in the cloud
- Choose a permission mode
- Configure the sandboxed Bash tool
- Checkpointing
- Set up Claude Code for your organization
- Track team usage with analytics
- Manage costs effectively
- Model configuration
- Security
- Data usage
- Zero data retention
- What certifications has Anthropic obtained?
- Ramp customer story
- Pictet turns weeks of work into hours with Claude Code
- Terminal-Bench 4.0 leaderboard
- Claude Code Reviews
- Claude Code Reviews & Ratings
- Claude Code
- Claude Code
- Anthropic unveils new rate limits to curb Claude Code power users
- Anthropic accidentally leaks Claude Code source via packaging error, no customer data exposed