Connect your AI to any API
UniversalBench lets your AI call any public API through one connection, with a key you store once and internal-network calls blocked for safety. It works with Claude, ChatGPT, Gemini, and any other MCP-compatible AI through one connection.
Connect any API in three steps
Get your API key
Get the key or token from the service you want to call.
Save it once in the secret vault
Store it in your UniversalBench secret vault under a name you choose, for example API_TOKEN.
Connect and ask
Your AI calls the API and references your key as a placeholder in the request header. The real value is substituted at call time and never reaches the model. Internal and private addresses are blocked.
What your AI can do
- ✓Call any public REST API your workflow needs
- ✓Store the API key once and reference it safely
- ✓Calls to internal and private addresses are blocked
Is there a any API MCP server to install?
No separate server to install or host. UniversalBench is the MCP server, and any API is one of the things it does for your AI behind a single connection. Save your credential once in the encrypted secret vault under API_TOKEN and it is used automatically after that.
The same connection also handles reaching the services your workflow depends on and more, so one setup covers far more than any API alone.
Frequently asked questions
How do I connect my AI to any API?
First, get the key or token from the service you want to call. Then save it in your UniversalBench secret vault under the name API_TOKEN, which unlocks the capability automatically. Finally, add the one MCP URL to your AI and ask in plain language.
Where do I put the credential?
In your UniversalBench secret vault, under the name API_TOKEN. It is encrypted, used automatically by the matching capability, and never shown back to you or sent to the model in plain text.
Does this work with ChatGPT and Gemini?
Yes. UniversalBench connects through one standard MCP link, so it works with Claude, ChatGPT, Gemini, and any MCP-compatible AI. You paste one URL and the capability is there.
