Claude Code

Claude Code

Claude Code inspects repositories, edits code, and runs commands from the terminal.

Oximy ResearchUpdated 19 September 2026

Overview

A
Transparency
7 / 10
Trust
94 / 100

Trains on your data

Free
Yes
Paid
No
Opt-out
Yes

Incidents

7
  • highHackers Stealing Claude Tokens from Subscribers
  • highAnthropic's AI Hacked Three Companies During Tests
  • mediumClaude Code: Sandbox Escape via Git Worktree Path Confusion Allows Unsandboxed Code Execution
SOC 2ISO 27001GDPRHIPAACSA STARISO 27017ISO 27018

Assessed SEP 21, 2026 · Checked by hand SEP 21, 2026

Commercial content is not used for training, consumer training can be disabled, qualified Enterprise accounts can use ZDR, and seven supported incidents remain.

Trust breakdown

A

94 / 100

Overall trust score

Data safetyComplianceIncidentsLegal
Data safety100
Compliance100
Incidents70
Legal100

Data and privacy

Trains on your data

Free
Yes
Paid
No
Opt-out
Yes

On by default. Admins can turn it off, which is why it costs less than training with no way out.

Storage regions

EuropeUnited StatesCanadaAsia-PacificJapanSouth KoreaSingaporeIndiaAustralia

Sub-processors

Google Cloud PlatformAmazon Web ServicesMicrosoft AzureCloudflareStripeWorkOSIntercomNutunBoldrTwilio

Data retention

Zero Data Retention (ZDR) for Claude Code is available to qualified accounts on Claude for Enterprise.

Deletion: You also are able to delete individual conversations, which will be removed immediately from your conversation history and automatically deleted from our back-end within 30 days.

Privacy policy

Incidents

Powered by

Claude Fable 5Claude Opus 4.8Claude Sonnet 4.6Claude Haiku 4.5
Proprietary model

Claude Code uses Claude models (Fable, Opus, Sonnet, Haiku) developed by Anthropic. It acts as an orchestration layer that wraps these models with tools for code interaction.

Integrations and access

Integrations

API: YesGitHubJIRASentryStatsigPostgreSQLFigmaSlackGmailTelegramDiscord

Available on

Web
iOS
Android
macOS
Windows
Linux
Extension
CLI

Pricing

Free

$0

Basic usage, included in the free tier of Claude.com

Pro

$17/mo

More usage, included in the Pro plan of Claude.com ($20/month if billed monthly, $17/month with annual subscription discount ($200 billed up front)). Perfect for short coding sprints in small codebases.

Max 5x

$100/mo

5x more usage than Pro, included in the Max plan of Claude.com

Max 20x

$200/mo

20x more usage than Pro, included in the Max plan of Claude.com. Great value for power users with the most access to Claude models.

Team/Enterprise

Custom

Claude Code is included in Team and Enterprise plans. Per-developer costs vary widely based on model selection, codebase size, and usage patterns such as running multiple instances or automation. Enterprise plans offer features like admin-set user and org spend limits, role-based access, SCIM, audit logs, compliance API, custom data retention, network-level access control, IP allowlisting, and HIPAA-ready offering. Claude Code also charges by API token consumption, with specific pricing for different models (e.g., Fable 5.1, Opus 5, Sonnet 5, Haiku 4.5) and a 'Fast mode' for Opus 4.8 at $30/$150 per million tokens.

Prices as listed on SEP 21, 2026; check the vendor's page.

Moat and openness

6/ 10
Moat 6 out of 10

Moat

  • Proprietary Model
  • Proprietary Data
  • Network Effects
  • Switching Costmoderate
  • Unique UX
  • Distributionstrong

Claude Code benefits from being built on Anthropic's proprietary Claude models, offering a unique orchestration layer and a comprehensive suite of developer tools. Its strong distribution through existing Claude plans and deep integrations create a moderate switching cost, but the underlying technology is still a general-purpose AI, and the market is highly competitive.

6/ 10
Openness 6 out of 10

Openness

  • Open model weights
  • Published research
  • Open source contributions
  • Transparency reports
  • Public safety evals

Anthropic has a strong commitment to transparency, evidenced by their dedicated Transparency Hub. They publish detailed information on model capabilities, safety evaluations, risk reports, and their Responsible Scaling Policy. They also engage with the research community and have a public bug bounty program. While they publish extensive research and transparency reports, there is no indication of open-sourcing model weights or significant open-source contributions.

Timeline

  1. 2026-02-05Opus 4.6 launched, and agent teams entered preview.
  2. 2025-12-08Claude Code integrated with Slack, allowing Slack mentions to start Claude Code web sessions.
  3. 2025-11-24Claude Opus 4.5 was released, expanding Claude Code's model options.
  4. 2025-10-31Plugins and marketplaces became a formal distribution system for Claude Code.
  5. 2025-10-20Cloud coding sessions for Claude Code entered research preview, making Claude Code available on the web.
  6. 2025-06-30Hooks, a feature for lifecycle-triggered automation, became available in the CLI.
  7. 2025-06-18Remote MCP servers became available, allowing Claude Code to gain OAuth connections to hosted MCP servers.
  8. 2025-05-22Claude Code moved out of research preview and became generally available. This announcement also introduced Claude Opus 4 and Claude Sonnet 4, models designed for coding, reasoning, and agent work. Native integrations for Visual Studio Code, JetBrains IDEs, and GitHub Actions were also announced.
Show all
  1. 2025-02-24Claude Code launched in research preview as a command-line tool for agentic coding, alongside the release of Claude 3.7 Sonnet. It was initially named Claude CLI.
  2. 2024-11Anthropic announced the Model Context Protocol, an open integration standard later adopted by Claude Code.

Company

Anthropic

Founded
2021
HQ
San Francisco, California, U.S.

Popularity

Not disclosed

monthly visits

Notable customers

StripeRakutenPwCMoney ForwardRamp

Claude Code holds a significant share of the AI coding market, with some reports indicating it commands approximately 54% of the market as of early 2026, outpacing competitors like OpenAI.

Value and ROI

What it should move: Engineers spend less time implementing changes, writing tests, and maintaining codebases.

Best fit

It fits teams with reliable tests, scoped permissions, and written repository rules.

The catch

Claude Code can read files and run commands. Limit access to repositories and secrets, set command permissions, and review changes before wider use.

  • Rakuten: Uses Claude Code for tests, API mocks, refactoring, bug fixes and documentation.[1]
  • Ramp: Uses Claude Code for application development, internal tools and incident investigation.[2]

In our stacks for

Engineering

References

  1. 01Rakuten: how it uses Claude CodeAnthropicVendor-published
  2. 02Ramp: how it uses Claude CodeAnthropicVendor-published
  3. 03Privacy policyAnthropicVendor-published
  4. 04Company siteAnthropicVendor-published
  5. 05SOC 2AnthropicVendor-published
  6. 06HIPAAAnthropicVendor-published
  7. 07ISO 42001AnthropicVendor-published
  8. 08FedRAMPAnthropicVendor-published
  9. 09Hackers Stealing Claude Tokens from SubscribersTechCrunchIndependent
  10. 10Anthropic's AI Hacked Three Companies During TestsAnthropicIndependent
  11. 11Claude Code: Sandbox Escape via Git Worktree Path Confusion Allows Unsandboxed Code ExecutionGitHub Advisory DatabaseIndependent
  12. 12Claude Code Sandbox Flaw Exposes Credentials and Source CodeCyberPress.orgIndependent
  13. 13Command Injection via Directory Change Bypasses Write ProtectionGitHub AdvisoryIndependent
  14. 14Malicious repo configuration can trigger data leakage via environment configuration used before trust confirmationGitHub AdvisoryIndependent
  15. 15Permissive Default Allowlist Enables Unauthorized File Read and Network Exfiltration in Claude CodeGitHub AdvisoryIndependent