How TokenHawk works
No SDK to install, no code migration: TokenHawk slots in front of your existing integration and does the work on the fly, on every request.
1. Create a TokenHawk key
From your dashboard, in 2 minutes. You provide your provider keys (OpenAI, Anthropic, Gemini, Mistral) there — they're stored encrypted, never accessible in plain text, not even by us.
2. Configure your code
Change your LLM client's base URL to TokenHawk, use your TokenHawk API key instead of the provider's, and prefix your models with their provider (e.g. openai/gpt-4o). That's everything that changes in your code.
# Before from openai import OpenAI client = OpenAI(api_key="sk-...") client.chat.completions.create( model="gpt-4o", messages=[...], ) # After from openai import OpenAI client = OpenAI( api_key="thawk_...", base_url="https://tokenhawk.aittow.com/api/v1", ) client.chat.completions.create( model="openai/gpt-4o", messages=[...], )
3. Full visibility
Every request is analyzed, sensitive data is masked or blocked, access is logged, and content is optimized before being forwarded to the provider.
The big picture
Works with your agent framework
TokenHawk exposes OpenAI's Chat Completions API — any client that speaks this format goes through the proxy by changing its base URL alone. Here's where the most common frameworks stand.
Use the ChatOpenAI client (langchain-openai) rather than ChatAnthropic or ChatGoogleGenerativeAI, which talk to the provider directly.
Route through the LiteLlm wrapper rather than a native Gemini model, which talks to Google directly.
Use OpenAIChatCompletionsModel explicitly — the SDK's default model (Responses API) isn't supported.
Native OpenAI-compatible client: add base_url and your TokenHawk key to your config_list.
CrewAI's LLM wrapper is already OpenAI-compatible: add base_url and your TokenHawk key.
Speaks Anthropic's native Messages protocol, not Chat Completions — not routable through TokenHawk today.
Full code snippets for each framework in your dashboard's Getting Started page.
Security & privacy
Your data. Your keys. Our responsibility.
Security isn't a feature — it's the foundation.
Your API keys are protected
Credentials are encrypted (AES-256-GCM) and handled with strict access controls.
Your prompts stay yours
We do not use your data to train models. Retention is minimal and configurable.
EU-based infrastructure
TokenHawk is operated from the EU, with data handling designed for European businesses.
Transparent by design
Clear documentation on what data is processed, stored, and for how long.
Test before you commit
Start safely, then scale.
You don't have to trust us with production on day one.
Test in a non-production environment
Use a test API key and your development environment. See exactly how TokenHawk works with your own traffic.
Move to production when ready
Once you're comfortable, switch to your production environment. No code changes required.
Frequently asked questions
Do I need to change my application code?
No. Only the base URL and API key change. Your prompts, logic and tools stay identical.
How long does integration take?
A few minutes: create a key, change an environment variable, redeploy.
Which LLM providers are supported?
OpenAI, Anthropic, Google Gemini and Mistral AI today, with the ability to switch providers without rewriting your integration.
Can I test before going to production?
Yes — create a test key separate from your production key. See our Security page for the recommended practice.