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

# Mid-Call Context Injection

> Send messages and instructions to your AI agent during a live call

## Overview

Context injection lets you send messages or instructions to your AI agent while a call is in progress. The agent receives the injected content as additional context and can act on it naturally within the conversation.

Use cases:

* Feeding the agent real-time information (e.g. order status lookups)
* Giving the agent new instructions mid-conversation
* Passing context from a human supervisor monitoring the call

## Request

```bash theme={null}
curl -X POST https://api.kejue.co/api/v1/calls/{call_id}/inject-context \
  -H "X-API-Key: kej_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "text": "The customer'\''s order #12345 shipped yesterday via FedEx.",
    "urgency": "soon"
  }'
```

| Field     | Type   | Required | Default  | Description                                     |
| --------- | ------ | -------- | -------- | ----------------------------------------------- |
| `text`    | string | Yes      | -        | The context or instruction to send to the agent |
| `urgency` | string | No       | `"soon"` | Controls when the message is delivered          |

## Urgency Levels

| Level       | Behavior                                                              |
| ----------- | --------------------------------------------------------------------- |
| `immediate` | Interrupts the current conversation to deliver the message right away |
| `soon`      | Delivers at the next natural pause or speaker turn                    |
| `later`     | Passive context — the agent incorporates it when relevant             |

<Info>
  Use `immediate` sparingly — it interrupts the natural flow. For most cases, `soon` provides the best balance of timeliness and conversational quality.
</Info>

## Response

Returns `204 No Content` on success.

## Error Handling

| Status | Condition                                        |
| ------ | ------------------------------------------------ |
| `404`  | Call not found in this workspace                 |
| `409`  | Call is not in `in_progress` status              |
| `400`  | Invalid urgency value                            |
| `502`  | Failed to deliver the message to the active call |

## Example: Real-Time CRM Lookup

<CodeGroup>
  ```javascript Node.js theme={null}
  const response = await fetch(
    `https://api.kejue.co/api/v1/calls/${callId}/inject-context`,
    {
      method: 'POST',
      headers: {
        'X-API-Key': 'kej_live_...',
        'Content-Type': 'application/json',
      },
      body: JSON.stringify({
        text: 'Customer has an outstanding invoice of $2,500 due in 3 days.',
        urgency: 'soon',
      }),
    }
  );
  // 204 = success
  ```

  ```python Python theme={null}
  import httpx

  response = httpx.post(
      f"https://api.kejue.co/api/v1/calls/{call_id}/inject-context",
      headers={
          "X-API-Key": "kej_live_...",
          "Content-Type": "application/json",
      },
      json={
          "text": "Customer has an outstanding invoice of $2,500 due in 3 days.",
          "urgency": "soon",
      },
  )
  # 204 = success
  ```
</CodeGroup>