DeepL for AI agents
Your agents can already translate. DeepL makes them translate the way your company does: your terms, your tone, your quality bar.
Available in Claude, ChatGPT, Microsoft Copilot, and any AI agent that supports MCP.
Choose how to connect DeepL
Copy the MCP URL or a command into your AI tool — your agent handles the rest.
Remote MCP
https://mcp.deepl.com/v1/mcp
Sign in once with your DeepL account — no API keys. For seat-based and Enterprise plans.
DeepL CLI
Translate text and files from the terminal — or let your coding agent run it for you.
Local MCP server
npx -y deepl-mcp-server
The open-source server — runs on your machine, uses your API key.
Anthropic connector
Pick DeepL in Claude's connector directory — certified by Anthropic, no MCP URL or setup.
Coming soon
Agent Skills
Prepackaged DeepL skills for Claude, Copilot, and other agent platforms — localization workflows your agents already know how to run.
Plugins
Native DeepL plugins for agent ecosystems and marketplaces — one-click install, no configuration.
Microsoft Copilot connector
Pick DeepL from Copilot's connector list — certified by Microsoft, no MCP URL or setup.
Works where your agents work
Claude
ChatGPT
Microsoft Copilot
Any AI agent
Why connect DeepL to your AI tools
DeepL is the language layer for your agents: one source of truth for terminology, style, and quality across every tool your team uses.
One source of truth, everywhere your agents work
The same DeepL translation, glossaries, and style across Claude, ChatGPT, Copilot, Word, and Outlook.
Apply your customizations just by asking
Ask in plain language — DeepL applies the right glossary, tone, and style automatically.
Enterprise-grade and certified
Your texts never train models, and admins control access. Certified for Microsoft Copilot and Claude.
Always know it's DeepL
Every translation is a real DeepL call — attributable, never a model guess.
Use cases
Localize without leaving the workflow
Attach a document and ask your agent to translate it. DeepL handles the languages; your files keep their layout.
- Layout and formatting preserved — tables, headers, and images stay in place
- Glossary terms applied across every file
Support customers in any language
Draft replies in all your customers’ languages at once — accurate, on-brand, and instant.
- Translate into many languages in one request
- Your support glossary keeps product names consistent
Put your glossaries to work
Your agent knows your terminology. Ask what glossaries you have, then apply them to any translation.
- List and use your glossaries straight from chat
- Terminology stays consistent in every language
Agents can translate, DeepL makes it production-grade
| Agent alone | With DeepL |
|---|---|---|
Consistency | Output varies with each run | Deterministic, the same result every time |
Your terminology | Re-explain your terms in every prompt | Glossaries and style set once, applied everywhere |
Documents | Content dropped or hallucinated | Whole files translated, layout preserved, no hallucination |
Cost at scale | Pays for the whole document in tokens, both ways | Pay for the source text, not tokens both ways |
Frequently asked questions
The Model Context Protocol (MCP) is an open standard that lets AI assistants securely connect to external tools. The DeepL MCP server gives any compatible assistant access to DeepL translation, writing, and glossaries — no custom integration needed.
Claude, ChatGPT, Microsoft Copilot, and developer tools like Cursor and Continue. Any client that supports remote MCP servers can connect to DeepL.
On a seat-based plan (Individual, Team, or Business) or DeepL for Enterprise, you simply sign in with your DeepL account when your AI tool asks — no API key needed.
You connect by signing in once with your DeepL account. If your organization uses single sign-on (SSO) with DeepL, that login goes through your identity provider as usual — no separate credentials to manage. Administrators control which users can connect at the organization level.
When you sign in with your DeepL account, MCP usage is included in your subscription and works within its fair-usage limits — there is no separate bill. The connector itself is free.
Yes. Requests are processed under your existing DeepL agreement, and your texts are never used to train models. Administrators can enable or disable agent access for the whole organization.