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

# telli

> Connect telli agents to company knowledge in Qontext

Connect telli to Qontext through a custom HTTP tool so your agents can search your context repository during calls.

With Qontext + telli, you can:

* Give your agents access to relevant, up-to-date company knowledge
* Maintain knowledge centrally in Qontext and use it across your agents
* Control which context each agent can access

***

## Before you start

You'll need a telli agent and a [Qontext API key](/api-docs/authentication#creating-an-api-key) with read access to the context you want the agent to use. Use a separate key for telli and configure its [Access Controls](/features/access-management#access-controls-for-api-keys) as needed.

***

## Installation

<Steps>
  <Step title="Add a custom tool in telli">
    Open your telli agent, go to **Tools**, and add a custom HTTP tool.

    | Setting | Value |
    | - | - |
    | **Name** | `search_qontext` |
    | **Description** | Search company knowledge to answer questions about our company. |
    | **HTTP Method** | `POST` |
    | **Response Timeout** | Start with `10` seconds for your first test. |

    Set **URL** to:

    ```text theme={null}
    https://api.qontext.ai/v1/search/hybrid
    ```
  </Step>

  <Step title="Configure the request parameters">
    Under **Parameters**, add these **Headers**:

    * **`Authorization`**: choose **Secret** and set the value to `Bearer <your-api-key>`.
    * **`Content-Type`**: choose **Constant** and set the value to `application/json`.

    Replace `<your-api-key>` with your Qontext API key, keeping `Bearer ` before it.

    Then add these **Body** fields:

    * **`query`**: choose **LLM Parameter** with type **String**. Use this description so telli generates a question from the conversation: "A self-contained search question based on the caller's request and relevant conversation context."
    * **`limit`**: choose **Constant** with type **Number** and set the value to `3`.

    Qontext returns matching knowledge for the agent to use in its answer. See [Search files by meaning](/api-docs/search/hybrid-search) for the API reference.
  </Step>

  <Step title="Tell the agent when to search Qontext">
    In telli's prompt editor, select the tool from the **Tools** sidebar to insert its reference. Specify when the agent should search Qontext, use the returned knowledge to answer, and acknowledge when it cannot find relevant information. Save the configuration.

    For more on telli's tool configuration, see [Custom HTTP Tools](https://docs.telli.com/custom-tools) and [Tool Calls](https://docs.telli.com/en/get-started/tool-calls).
  </Step>
</Steps>

***

## Test the connection

Start a test call and ask a question whose answer is in Qontext but isn't already in the agent's prompt.

In Qontext, open the telli API key's [Retrieval logs](/features/retrieval-logs). Check that the call generated a search with a relevant question and returned the expected knowledge. Confirm that the agent used that knowledge in its spoken answer.

<div className="qontext-tip-check-wrapper">
  <Check>
    Once the search returns the expected knowledge and the agent uses it in its answer, your telli agent is connected to Qontext.
  </Check>
</div>


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