Anthropic's agentic coding tool for the terminal. Autonomous file edits, test runs and multi-step refactoring in existing codebases.
Claude Code is the official agentic coding tool from Anthropic, the company behind the Claude model family. Unlike classic AI autocomplete assistants like GitHub Copilot, Claude Code runs as a standalone agent in the terminal and executes multi-step engineering tasks autonomously: it reads files, executes bash commands, changes code, runs tests, interprets errors and iterates until the goal is reached.
The tool has been generally available since 2025 and is part of the official Claude Code ecosystem, which alongside the CLI includes plugins, skills, hooks and an MCP server framework. Claude Code primarily uses Claude Sonnet for fast tasks and Claude Opus for complex planning — model selection is controllable via flag or environment variable.
Particularly relevant for the DACH market: Claude Code can run via AWS Bedrock in EU region hosting (Frankfurt / eu-central-1). Combined with the Anthropic Enterprise DPA, zero-data retention and a clean logging classification, a GDPR-compliant deployment is possible — but only with deliberate setup, not out-of-the-box.
Runs entirely in the terminal. No IDE binding, works in tmux, over SSH, on VPS, in containers.
Read, Write, Edit, Bash, Grep, Glob — Claude Code uses these tools autonomously and orchestrates them via multi-step plans.
Model Context Protocol exposes external services (Supabase, GitHub, Browser-Use, Slack) as tools. Extensible without client update.
Plan mode shows intent before code is changed. Permissions can restrict bash whitelist, write scopes and hook checks.
Reusable skills encapsulate domain knowledge (invoicing, legal, SEO). Hooks trigger automation before/after tool calls.
AWS Bedrock + Google Vertex AI as backend supported. EU region, DPA, zero retention — all legally orchestrable.
The DACH region is a specific market for Claude Code: engineering teams are technically strong, but compliance requirements (GDPR, DPA, AI Act) slow broad enterprise rollout. Companies wanting to deploy Claude Code in Germany, Austria or Switzerland have two options:
For companies with around 50+ developers, the agency path typically pays off — the discovery phase is cleaner, and the DPA process with Anthropic Customer Success is non-trivial. Smaller teams can introduce Claude Code themselves — the core CLI is well documented.
An important point: Claude Code is a US product even when hosted via EU regions. For government, healthcare and financial contexts, sovereignty questions remain. Here a comparison with self-hosting-capable alternatives like OpenClaw, which runs locally and supports a strict GDPR pattern, is worthwhile.
A quick positioning of the four dominant agentic coding tools in 2026.
| Tool | Interface | Model control | Self-host | Best use-case |
|---|---|---|---|---|
| Claude Code | CLI | Anthropic-hosted (EU via Bedrock) | Nein | Multi-step engineering in existing codebases |
| Cursor | IDE | Multi-Model (Anthropic, OpenAI) | Nein | Inline coding + chat for individual developers |
| Cline | VS Code Plugin | Multi-Model (BYO-API) | Via Ollama yes | OSS agentic coding in VS Code |
| Aider | CLI | Multi-Model (BYO-API) | Via Ollama yes | Git-commit-centric OSS workflows |
Five steps for a compliant enterprise setup. This list is also annotated as HowTo schema on the page — search engines read it directly as a structured guide.
Sign DPA with Anthropic — enterprise account, DPA request via customer success, document sub-processor list.
Pin EU region via AWS Bedrock — switch model routing to eu-central-1 (Frankfurt).
Enable zero-data retention in the dashboard — prevents training on your prompts.
Define logging classification — pseudonymised logs, PII filter before storage.
Onboard team + write prompt policy — what is allowed, what not, integrate CI check.
Note: These steps do not replace legal advice. Before enterprise rollout, have a data protection legal review performed.
Claude Code is Anthropic's official agentic coding tool. It runs as a command-line assistant in your terminal and works directly on an existing codebase: file edits, test runs, bash commands and multi-step refactoring are executed autonomously. Unlike pure autocomplete tools like GitHub Copilot, Claude Code operates with tool-use, plan-execution and a large context window.
Claude Code is the client layer and uses the Claude model family from Anthropic — primarily Claude Sonnet and Claude Opus depending on the plan. Model selection happens per flag or ENV; the workflow stays identical. For sensitive workloads, Claude Code can also run against AWS Bedrock or Google Vertex AI — region hosting is then centrally controllable.
Yes — with the correct setup. Anthropic offers Enterprise contracts with DPA (Data Processing Agreement), EU region hosting via AWS Bedrock and zero-data retention on the API layer. Companies rolling out Claude Code GDPR-compliant need a discovery phase for DPA, region pinning, logging classification and sub-processor listing.
Claude Code is CLI-first and focused on agentic workflows in existing codebases — ideal for engineers who live in the terminal. Cursor is an AI-first IDE with inline completions and chat. Cline is a VS Code plugin with a similar agent approach as Claude Code. Aider is CLI-first with git-commit integration. The choice depends on team workflow — a specialised agency knows all four and advises per use-case.
Claude Code as a CLI tool is free. API usage is billed: Claude Sonnet is approximately $3/MTok input + $15/MTok output (as of 2026-04). Teams get Claude Pro (~ $20/month) and Claude for Business with centralised billing + DPA. Agency implementation (setup, workflows, MCP servers, onboarding) adds 8,000-25,000 EUR for an MVP.
Trusted AI Partners lists verified DACH agencies using Claude Code in production — with real references, not just demos. Filter by Claude Code and Anthropic in the tech stack. For specific searches (agency, freelancer, city drill-down) use the dedicated landing pages under /en/tools/claude-code-agency and /en/tools/claude-code-freelancer.
Describe your project — we match three suitable DACH providers within 24 hours.