Skip to main content

Connect a model provider

Agents cannot run without a model provider. Connecting one is what makes models selectable on an agent and lets the platform perform its own built-in AI tasks.

Provider configuration lives in the Settings section under Administration.

Important

With no provider configured, every model in an agent's Model list is locked, with the tooltip "This model requires the Creatio GenAI proxy. Configure it at Settings → Creatio Generative AI." Every run then fails with an LLM provider error, no matter how well the agent is configured.

Connect Creatio Generative AI​

Creatio's managed AI service is the quickest route to a working provider.

  1. Go to the Settings section under Administration in the navigation panel.

  2. Open the Creatio Generative AI page.

  3. Turn on Use Creatio Generative AI.

  4. Fill out the connection fields.

    Field

    Field value

    Creatio AI Service URL

    The address of the service, for example, https://ai.example.com.

    Identity Provider URL

    The identity provider that issues the token, for example, https://identity.example.com.

    Client ID, Client Secret

    The OAuth credentials. Leave the secret empty when editing to keep the existing value.

    All four fields are required. Testing with any of them empty returns "Connection failed: Creatio LLM Proxy is not fully configured (missing proxy URL, identity URL, client ID, or client secret)."

  5. Click Test Connection. Creatio AI Studio reports "Connection successful! Token acquired." when the credentials work.

  6. Click Save changes.

As a result, agents can make LLM calls through the managed service. Every model picker lists the Creatio-provided models first, under Creatio provided, while a model your organization serves with its own key is listed under its vendor.

note

Test the connection before saving. A configuration that saves cleanly can still fail to acquire a token, and the failure then surfaces as a runtime error on an agent rather than a setup error here.

Set the Platform LLM​

The Platform LLM page sets the model behind each built-in AI task, separately from the model an individual agent selects. There is no single platform-wide model: each task is a card with its own Model picker and its own Save.

  1. Open the Platform LLM page in Settings.

  2. Set Default model. Every agent set to "Use platform default" uses this model, including existing agents, so changing it can affect how they work. An agent with a specific model stays on it.

    If the new model cannot do something that some agents rely on, Creatio AI Studio lists those agents in the "Some agents will lose a capability" dialog. Click Change anyway to continue, or Cancel to keep the current model.

    When no model is selected, "Use platform default" applies, which is GPT-5.6 Terra, a Creatio-provided model. GPT-4o cannot be selected, but an organization whose default is already GPT-4o keeps receiving it until the setting is changed.

  3. Set the model for each remaining task. A card you leave unset uses the fallback it names.

    Card

    What it does and what applies when unset

    Conversation titles

    Writes a short title for each new AI Twin conversation from its first message. When no model is selected, the card shows GPT-5.6 Luna, and "Use Luna (default)" returns the card to it.

    Approvals

    Writes the confirmation prompts shown in a conversation. When no model is selected, "Use agent's LLM (default)" applies, so the agent's own model writes them.

    Memory search planning

    Turns each request into the searches the agent runs against its memory. When no model is selected, "Use memory runtime default" applies. If the selected model cannot be used, the agent's own model plans the search.

    Speech-to-text

    Transcribes the messages people dictate in the Preview panel of an agent. When no model is selected, the platform default applies. Dictation is recorded in usage with its cost in AI credits.

  4. Click Save on each card you changed.

As a result, each built-in AI task runs on the model you selected for it.

Other provider keys​

Page

What it configures

Voice provider keys

Credentials for the speech providers a voice channel uses, and for avatar providers. Avatar support is enabled per organization and appears only where it has been turned on. Learn more: Add a Voice channel.

Credentials store

Reusable OAuth and bearer credentials referenced by integrations and CRM connections. Learn more: Connect a CRM per environment.

See also​

Create a prompt agent

Connect a CRM per environment

Create an API key