Integrations

Your models, your engines, your rules.

For each building block of the platform, you choose your provider: on your premises, in a European cloud or with a major vendor, and you can switch without moving your data or reconfiguring your agents. For each one, here is who runs it, where it runs and how it is billed.

AI models

On-premise or in the cloud, with your own key: you contract directly with the provider, at the price you negotiated. SmartAGT takes no margin on inference.

Connectors

Agents work inside your business tools, with the rights of the connected account. Every write action can require human approval.

Web search

For agents that need to consult the web, with a domain allow-list if you want one.

Vector databases

Where your documents are indexed for the agents' search: inside your infrastructure or in a managed service.

Our method

How we integrate a provider

Plugging in an API is not enough: a provider has to work with the platform's agents, tools and quotas. Every integration follows the same path.

Connection

Plugged into our shared OpenAI-compatible transport, which does most of the connecting work.

Qualification

Capabilities checked offer by offer: chat, streaming, tool calls, embeddings.

Real-world testing

Real calls from the platform, never on the documentation's word alone.

Price lists kept in step

Provider price lists are published in a reference set your platform can synchronise, so the cost you see tracks the real one.

Your provider is not on the list? Any server or service exposing the OpenAI-compatible API can be connected: let's talk.

Further reading

MODEL-AGNOSTIC BY DESIGN

Want to see your models
plugged into your data?