Dashboard
Credits remaining, monthly consumption, total requests at a glance.
One API.
Every model.
GDPR-compliant.
OpenAI-compatible as a drop-in replacement. With anonymization per parameter.
The anymize developer console is your control center for API integrations: one API key for every frontier model (GPT, Claude, Gemini, Mistral, Perplexity, Kimi) plus our own models (Waterfall, Fountain). Supports the OpenAI schema – swap the base URL and your existing code keeps running. With anonymization when you need it. With fallback models when a provider fails. With Zero Data Retention when your compliance team demands it.
What you get
Everything in
one dashboard.
The developer console bundles every management function you need for productive API use.
API keys
Create, name, rotate and revoke multiple keys in parallel.
Models
Every available model with price per 1M input/output tokens.
Chat playground
Test models live with system prompt, temperature, max tokens, top p.
Configuration
Fallback models, Zero Data Retention – one toggle per feature.
Usage
Cost tracking, rate limits, complete logs of every request.
All reachable via Back to chat – the console is not a separate product but the switch next to your normal anymize work.
OpenAI-compatible
Swap the base URL.
That's it.
The anymize API implements the OpenAI schema for chat completions. Any code that speaks to OpenAI today – the official Python SDK, the Node SDK, LangChain, LlamaIndex, your own backend or a no-code tool – works against anymize the moment you swap two values.
Before · OpenAIpython
1from openai import OpenAI2client = OpenAI(4api_key="sk-proj-...",Changed5base_url="https://api.openai.com/v1"Changed6)
After · anymizepython
1from openai import OpenAI2client = OpenAI(4api_key="anymize_...",Changed5base_url="https://app.anymize.ai/api/v1/llm"Changed6)
No rewrite, no new SDK dependency, no parser overhaul. The chat/completions endpoint schema is identical – same roles, same messages, same parameters (temperature, max_tokens, top_p, stop, etc.). Streaming is supported.
What differs: the model names. Instead of "model": "gpt-5" at OpenAI you address anymize identifiers ("openai/gpt-5", "anthropic/claude-opus-4.7", "google/gemini-2.5-flash", "waterfall-1.0", "fountain-1.0"). The current list you can query at any time via GET /models.
Anonymization per parameter
The feature that
no other API provider has.
The exciting trait of the anymize API: you enable anonymization via a parameter in the request. For every single call you decide whether the content runs through our anonymization pipeline before it reaches the chosen model.
With anonymization
For sensitive data
You want to use a frontier model (GPT, Claude, Gemini) but your prompt contains personal data. If you enable anonymization, anymize replaces names, addresses, IBANs, case numbers and 40+ more categories before sending. The model's answer is automatically retranslated to your original data before it is returned. From your application's point of view the API behaves like a regular OpenAI API – only without any third-country transfer of personal data.
Without anonymization
For non-critical content
For content without personal reference – public studies, internal handbooks, structured data – you skip the anonymization step. Direct route, lower latency.
Bonus option
Own models
If you use our own models Waterfall or Fountain, you don't need anonymization in the first place: the models run with us in the EU, the data never leaves the EU. Ideal for code, Excel and JSON work, where anonymization would destroy semantic value.
The exact parameter syntax is in the API documentation – either via the dedicated endpoint for anonymized calls or as a body parameter on the standard endpoint, depending on your integration preference.
API keys
Multiple keys.
Granular control.
A single key is rarely enough. Production apps, staging environments, internal tools, ad-hoc scripts – for every setup a dedicated key, nameable, deletable, rotatable.
console · api keys
Your API keys
3 keys · 1 actively in use
anyclaw
anymize_jv•••••jPJk
CreatedMar 12, 2026
Test-Staging
anymize_kq•••••9xTz
CreatedFeb 27, 2026
n8n-Workflow
anymize_pf•••••aB12
CreatedJan 4, 2026
Why multiple keys matter
Isolation on leak
A compromised key is revoked without paralyzing the rest.
Traceability
Per-key logs show which integration is making which requests.
Rotation hygiene
Regular rotation is a standard recommendation in audits (GDPR, ISO 27001).
Models, prices, credits
One wallet
for everything.
The models overview in the console shows three groups, each model with input/output price per 1 million tokens – toggleable between EUR and credits.
anymize's own models
European inference. No external providers, no additional contracts.
Deep thinking
anymize/waterfall-1.0
Input · 1M
€0.60
Output · 1M
€3.00
Performance
anymize/fountain-1.0
Input · 1M
€0.15
Output · 1M
€0.60
EU models
Variants hosted in Europe – for applications where the EU hosting aspect matters.
Gemini · EU
google/gemini-2.5-flash-eu
Input · 1M
—
Output · 1M
—
Mistral · EU
mistral/large-eu
Input · 1M
—
Output · 1M
—
GPT · EU
openai/gpt-5-eu
Input · 1M
—
Output · 1M
—
International models
Claude, GPT, Gemini, Perplexity, Kimi at full strength. For sensitive data: enable anonymization.
OpenAI
openai/gpt-5
Input · 1M
—
Output · 1M
—
Anthropic
anthropic/claude-opus-4.7
Input · 1M
—
Output · 1M
—
Google
google/gemini-2.5-pro
Input · 1M
—
Output · 1M
—
Perplexity
perplexity/sonar-pro
Input · 1M
—
Output · 1M
—
Moonshot AI
moonshot/kimi-k2
Input · 1M
—
Output · 1M
—
One credit wallet for everything
One plan. One budget. One billing view.
The specialty: credits are universal. From the same wallet you pay for:
Chat usageAnonymizationAPI callsKnowledge basesArtifact generationLive transcription
No separate API subscriptions on top of chat subscriptions. Frontier models run at provider price plus a thin anymize margin.
Chat playground
Test before
deploy.
The playground is a reduced chat surface built for developer workflows – not for end-user content. You test exactly what your API calls would produce. Configuration goes one-to-one into the API body of your application.
You are a precise assistant for legal documents. Answer in the language of the input, structured, only based on the supplied file…
anymize/waterfall-1.0
Temperature0.3
Max tokens4,096
Top P0.9
playground · session
—
Enter a message…
Especially useful for
Prompt engineers
Tune system prompts iteratively with the real parameters that later run in production.
Developers
Evaluate a new model candidate before it goes into production.
Ops teams
Reproduce an incident from a production API call – with the same parameters.
Fallback & Zero Data Retention
Reliability and data hygiene –
toggleable in the console.
Under configuration sit two enterprise features that many API providers do not offer at all.
Fallback models
Answer guaranteed.
If your primary model does not answer in time (timeout, provider outage, rate limit), a fallback model kicks in automatically. You configure two in series – the second is often an in-house model, independent of external providers.
Primary
openai/gpt-5
Default role
→
Fallback 1
anthropic/claude-opus-4.7
Same performance class
→
Fallback 2
anymize/waterfall-1.0
Independently available
Zero Data Retention
One toggle, two behaviors.
A toggle decides how anymize treats your request data. Metadata (tokens, model, timestamp) remain in both modes – content only when required.
ZDR off
Documents and content are stored in your account for later inspection (logs, reuse).
ZDR on
Request content is cleaned up after processing. Ideal for legal, healthcare, finance.
For industries with strict compliance requirements, ZDR is often set – the platform then behaves largely like a pass-through proxy with anonymization and fallback logic.
Usage tracking
Every request
traceable.
Under usage you find three sub-areas – for budget, scaling and forensics.
Costs
Monthly and daily breakdown in EUR or credits, filterable by key, model and time range. You see immediately: which key burns the budget, which model is the cost driver, which workflow is inefficient.
Rate limits
Your current limits (depending on your plan) at a glance – plus history of how often you actually hit them. Anyone scaling productively sees in time whether an upgrade is due.
Logs
Every API request lands in the log: timestamp, API key, model, input/output tokens, latency, status. For debugging, audit, compliance evidence – and for the question “did our software really send it like that?”.
With Zero Data Retention enabled, metadata remains in the log, content is removed – billing is still complete.
Use cases
Where developers
deploy the anymize API.
Six typical integration scenarios.
Drop-in replacement for OpenAI
What happens
Base URL swapped, rest of the codebase unchanged
Why anymize
GDPR-compliant via anonymization, every model available, one contract
n8n / Make.com / Zapier
What happens
No-code workflows with AI steps
Own backend (Node / Python / Go)
What happens
Standard LLM requests from your application
Why anymize
Fallback models, ZDR, central credit wallet
Agentic workflows (LangChain, LlamaIndex, MCP)
What happens
Tool calling, RAG, agents with multiple LLMs
Why anymize
Multiple models with one key, cost tracking per tool call
CRM / DMS integration
What happens
Automated summaries, draft generation
Why anymize
Anonymization protects client/customer data from US providers
Dev tools (IDE plugin, internal code assistant)
What happens
Code review, refactoring, documentation generation
Why anymize
Waterfall/Fountain for code without anonymization – European stack
What you should know about the developer console.
Frequently asked questions
Yes. The chat completions endpoint implements the OpenAI schema. You swap base_url and api_key in your existing code – that's it. The official OpenAI Python SDK, the Node SDK, LangChain, LlamaIndex, Claude Code, Aider, Cursor and anything else that supports the OpenAI format works immediately. Only the model names change to the anymize identifiers (e.g. openai/gpt-5, waterfall-1.0).
We stand behind anymize. And we know – when an AI tool touches client, patient or employee data, a demo video isn't enough. That's why we give you 14 days of full access – all models, all features, no credit card. Enough time to be certain, before you trust us.
Your AI workplace awaits.