> ## Documentation Index
> Fetch the complete documentation index at: https://docs.callers.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Bring Your Own LLM

> Run your Callers agents on your own large language model endpoint or provider account, for compliance, existing contracts, or custom models.

Enterprise customers can run Callers agents on their own large language model (LLM) instead of Callers' default models. Your agents keep working the same way, with the same scripts, actions, knowledge base, guardrails, and post-call analysis, but the conversation is generated by a model endpoint or provider account that you control.

> [!NOTE]
>
> Bring your own LLM is available on enterprise agreements and is configured together with the Callers team. There's no self-serve setting in the dashboard.

## Why bring your own LLM

* **Data residency and compliance.** Keep model processing in a region, cloud, or provider account approved by your security and legal teams.
* **Existing provider contracts.** Use the commitments, rate limits, and terms you already have with a model provider.
* **Fine-tuned or private models.** Use a model trained on your domain, terminology, or tone.
* **Cost and usage control.** Track and manage model usage in your own provider account.

## How it works

Callers keeps handling telephony, speech recognition, voices, turn-taking, actions, and analytics. Only the model that generates the agent's replies changes.

## What to prepare

Real-time voice conversations put strict demands on the model. Before you contact us, gather:

| Item | Details |
| - | - |
| **Provider and model** | Which provider or self-hosted stack you use, and which model or deployment. |
| **Endpoint and credentials** | How Callers will reach the model, and how access is authenticated. Plan to issue dedicated credentials for Callers. |
| **Region** | Where the model runs, and any data-residency requirements. |
| **Expected volume** | Peak concurrent conversations and monthly conversation volume, so we can confirm your capacity and rate limits. |
| **Latency** | Typical time to first token from the regions where your callers are. |
| **Capabilities** | Whether the model supports streaming responses and tool (function) calling. |

> [!WARNING]
>
> Voice agents need **low-latency streaming** to avoid awkward pauses, and **tool or function calling** to run actions such as transfers, bookings, and HTTP requests during the call. A model without these may not work well, or at all, for voice. Text and email agents are less sensitive to latency.

## Get started

**1. Talk to us about your provider**

[Book a meeting](https://www.callers.ai/book-a-meeting) or email [support@callers.ai](mailto:support@callers.ai). Share the details from [What to prepare](#what-to-prepare).

**2. Confirm compatibility**

The Callers team reviews your model and endpoint and confirms whether they fit your use case.

**3. Connect and test**

We connect your model to your account. Test your agents with [test calls and chats](https://docs.callers.ai/documentation/build-your-agent/testing-your-agent.md) and compare the results with your current setup.

**4. Roll out**

Once you're happy with quality and latency, move live traffic to the agents that use your model.

## FAQ

> **Which providers do you support?**
>
> Talk to us about your provider. Compatibility depends on the model's latency, streaming, and tool-calling support, so we confirm it case by case.

> **Do I have to change my agents?**
>
> Scripts, actions, and settings stay the same. Different models can behave differently, though, so test your important flows and adjust prompts and instructions if needed.

> **Can I use my own model for some agents only?**
>
> Discuss your rollout plan with the Callers team during setup.

## Related

> **[Callers for Enterprise](https://docs.callers.ai/documentation/enterprise/overview.md)**
>
> All enterprise capabilities.

> **[Security and privacy](https://docs.callers.ai/documentation/account/security-and-privacy.md)**
>
> Data retention and sensitive data controls.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.