Connect a custom LLM to Creatio.ai
This functionality is available for Creatio version 8.3.2 and later.
You can connect your own large language model (LLM) to Creatio.ai, for example, if you use a proprietary LLM in your company. After you connect an LLM, Creatio.ai can use it in AI Skills, AI Agents, and system actions.
The following models can be connected in a streamlined fashion:
- OpenAI
- Azure OpenAI
- Custom LLM supported by LiteLLM. Learn more in the official LiteLLM documentation: Providers.
You can connect other models as well, but this requires additional setup in JSON.
Minimum LLM requirements
For API-based AI Skills and lightweight scenarios:
- сontext window of at least 25,000 tokens
- support for structured output
For chat and AI Agent scenarios:
- сontext window of at least 40,000 tokens
- support for function calling
- stable long-context performance
Connect an OpenAI model
-
Open the AI command center workplace → Creatio.ai models.
-
Click New → OpenAI.
-
Fill out the fields on the model page:
Field
Field value
Name
Human-readable model name displayed in the Creatio UI.
Description
Description of intended use cases. Optional.
Model name
The exact technical model ID from OpenAI, for example,
gpt-5.2-2025-12-11.API key
API key from OpenAI.
-
Click Save.
As a result, the OpenAI model will become available for use in Creatio.
Connect an Azure OpenAI model
-
Open the AI command center workplace → Creatio.ai models.
-
Click New → Azure.
-
Fill out the fields on the model page:
Field
Field value
Name
Human-readable model name displayed in the Creatio UI.
Description
Description of intended use cases. Optional.
Model name
Full Azure OpenAI model name.
API key
API key from Azure AI.
Azure resource name
Azure resource ID.
API Version
API version the resource uses.
-
Click Save.
As a result, the Azure OpenAI model will become available for use in Creatio.
Connect a custom LLM
Custom LLMs are models that:
- are not listed among the default supported LLMs
- run in your own infrastructure
- are hosted by a third-party provider
To connect a custom LLM:
- Open the AI command center workplace → Creatio.ai models.
- Click New → Custom.
- Specify the Name and Description.
- Provide the connection configuration in JSON format.
- Click Save.
As a result, the custom model will become available for use in Creatio.
JSON configuration examples
Cerebras (gpt-oss-120b)
{
"model": "cerebras/gpt-oss-120b",
"api_key": "YOUR_API_KEY"
}
Google Gemini 2.5
{
"model": "gemini/gemini-2.5-pro",
"api_key": "YOUR_API_KEY"
}
Local or unsupported providers
For models deployed on-premises or hosted by providers not supported by LiteLLM natively, if such models have OpenAI compatible API, for example, Mistral:
{
"model": "openai/mistral",
"api_key": "YOUR_API_KEY",
"api_base": "http://0.0.0.0:4000"
}
where:
model: the model name starting withopenai/.api_key: the API key if the server requires authentication.api_base: the local API endpoint.
Optional generation parameters
You can add model-specific parameters, such as:
temperaturetop_pmax_tokensmax_completion_tokensdrop_params
Example:
{
"model": "openai/mistral",
"api_key": "YOUR_API_KEY",
"api_base": "http://0.0.0.0:4000",
"temperature": 0,
"max_tokens": 50000
}
Use the LLM in AI Agents and AI Skills
- Open the AI command center workplace → Creatio.ai agents or Creatio.ai skills.
- Open the relevant AI Agent or AI Skill.
- Select the required LLM in the Model field.
As a result, the AI Agent or AI Skill will use only the selected LLM.
Set a default LLM for system actions
Creatio.ai uses a default LLM for system actions that are not tied to a specific AI Agent or AI Skill. To change the default LLM:
-
Click
to open the System Designer. -
Go to the System setup block → Lookups.
-
Open the LLM Model Categories lookup.
-
Click New → specify the record parameters.
Field
Field value
Name
Any human-readable name.
Code
Default.Default LLM Model
Select the needed model.
-
Click
.
As a result, Creatio will use this model in the following cases:
-
first user messages without a selected Skill or Agent
-
AI Agents or AI Skills that do not have an explicitly defined model
-
system actions, including the following:
- chat title generation
- response validation and cleanup
- background LLM operations