Limbik

Limbik is your AI resonance layer—predicting how real audiences interpret and respond to messages, concepts, and AI outputs before they reach the market. Powered by continuous primary research across 60+ countries and 25+ languages, Limbik delivers human-validated synthetic audiences—digital populations that simulate real audience response at machine speed and with research-grade accuracy (95% confidence, 1.5% to 3% margin of error). Limbik gives you the ability to immediately ensure your messaging resonates with what your target audience believes and feels.

This integration is maintained by Limbik.

Prerequisites

The following are required to use Limbik with Braze:

Before using any of the forecast endpoints, you must first identify which organization (account_id) you have access to. While most customers have only one organization, some accounts may have multiple organizations available.

1
2
3
curl -X 'GET' \
  'https://cortex.prod.limbik.com/rest/api/organizations' \
  -H 'accept: application/json'

1
2
3
4
5
6
7
8
{
  "data": [
    {
      "uid": "aca61bd5-7132-499c-946e-42d092cc1156",
      "name": "Braze API"
    }
  ]
}

Authentication

To access the API endpoints, you need a bearer token for authentication. Obtain your token by authenticating with your credentials.

1
2
3
4
5
6
7
8
curl -X 'POST' \
  'https://cortex.prod.limbik.com/rest/api/auth/login' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "username": "your_username",
  "password": "your_password"
}'

1
2
3
4
{
  "access_token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...",
  "token_type": "Bearer"
}

1
Authorization: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...

Use case - Generating message copy

By using both Braze and Limbik’s REST API endpoints, you can use Limbik’s generative forecasts to create message copy and send it through Braze messaging channels, or adjust existing copy to improve impact for your audience. Both platforms expose functionality you can call programmatically to build sophisticated workflows.

This documentation outlines two examples: generating message copy in Limbik and using this copy in a subsequent message sent through Braze, as well as using Limbik to score the quality of a given message for your chosen audience.

1
2
3
4
5
curl -X 'GET' \
  'https://cortex.prod.limbik.com/rest/api/forecasts/generate/template?prompt=YOUR_PROMPT' \
  -H 'account_id: YOUR_ACCOUNT_ID' \
  -H 'Authorization: Bearer YOUR_ACCESS_TOKEN' \
  -H 'accept: application/json'

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
[
  {
    "type": "Message",
    "displayText": "Formula one next race",
    "additionalDetail": "The latest dev in Formula...",
    "messages": [
      {
        "body": "The latest dev in Formula ..."
      }
    ],
    "population": {
      "id": 56,
      "name": "us2",
      "org_enabled": true,
      "org_visible": true,
      "categories": [],
      "display_name": "US Adults",
      "composite_key": "us2",
      "enabled": true
    }
  }
]

Response fields

The response contains the following key fields:

• type: The message type (for example, "Generate" for AI-generated content, "Message" for validated messages)
• displayText: A short title or summary of the message
• additionalDetail: The complete AI-generated message copy - This is the primary field containing the full message text that you can send through your messaging platform
• population: The target population and segments for this message

Using with Braze

The additionalDetail field from Limbik’s response contains the message copy you send to Braze. One common integration pattern is to pass that value in the trigger_properties.payload when calling the Braze trigger send endpoint. In the following example, replace with the actual string from Limbik's `additionalDetail` field, and replace with your campaign ID.

Braze trigger message request example

1
2
3
4
5
6
7
{
  "campaign_id": "",
  "trigger_properties": {
    "payload": ""
  },
  "broadcast": true
}

Use case - Synthetic audience details

To build on the first use case, use Limbik’s endpoint /rest/api/populations/{account_id}/{population_id}.

This endpoint returns key data points that describe the makeup of Limbik’s synthetic audiences, such as gender, location, and so on. You can use these values to populate Connected Audience objects when calling Braze’s messaging endpoints.

To obtain forecast scores for specific segments, identify the available countries and their corresponding segments.

Step 1: List available countries

Retrieve the list of countries available for your account:

1
2
3
curl -X 'GET' \
  'https://cortex.prod.limbik.com/rest/api/populations/list/aca61bd5-7132-499c-946e-42d092cc1156' \
  -H 'accept: application/json'

From the response, identify the country you want to use. For example, the United States has an id of 56.

Step 2: Retrieve available segments

After you retrieve the country ID, retrieve the full list of segments for that country.

1
2
3
curl -X 'GET' \
  'https://cortex.prod.limbik.com/rest/api/populations/aca61bd5-7132-499c-946e-42d092cc1156/56' \
  -H 'accept: application/json'

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
[
  {
    "id": 56,
    "name": "us2",
    "composite_key": "us2",
    "categories": [
      {
        "id": 9331,
        "name": "gender",
        "composite_key": "us2::gender",
        "segments": [
          {
            "id": 63793,
            "name": "female",
            "composite_key": "us2::gender::female"
          }
        ]
      }
    ]
  }
]

1
2
3
4
5
6
7
8
9
10
11
{
  "AND": [
    {
      "custom_attribute": {
        "custom_attribute_name": "education_level",
        "comparison": "equals",
        "value": "masters"
      }
    }
  ]
}

Use case - Evaluating forecast score

You can use Limbik to create an estimated score for a message against a synthetic audience. Do this programmatically with Limbik’s forecasts/synchronous endpoint.

Option 1 - Synchronous forecast

You can use the response payload from the template generation directly with the synchronous forecast endpoint:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
curl -X 'POST' \
  'https://cortex.prod.limbik.com/rest/api/forecasts/synchronous' \
  -H 'accept: application/json' \
  -H 'account_id: aca61bd5-7132-499c-946e-42d092cc1156' \
  -H 'Content-Type: application/json' \
  -d '{
  "type": "Generate",
  "displayText": "Formula one season testing 2026",
  "additionalDetail": "Day 1 of the 2026 Formula 1 Bahrain testing session has concluded. Lando Norris recorded the fastest time in the McLaren, with Ferrari in second place. Cadillac drivers Sergio Perez and Valtteri Bottas completed 107 laps, nearly two race distances, and Audi introduced significant upgrades. Which team do you expect to perform best in Australia? #F12026 #BahrainTesting #LandoNorris",
  "population": {
    "population": "us2",
    "segments": []
  }
}'

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
{
  "uid": "6c5e28ef-8796-4659-a743-d842a06c9bf7",
  "datetime": "2026-02-11T20:04:06.545+00:00",
  "userId": "9cdd921c-f62f-46a6-902f-a6b0d1702f99",
  "accountId": "aca61bd5-7132-499c-946e-42d092cc1156",
  "name": "Formula one season t...",
  "user_message_context": "",
  "population": [
    {
      "name": "us2",
      "display_name": "US Adults",
      "categories": []
    }
  ],
  "privacy_compliant": false,
  "model_outputs": {
    "belmetrics": {
      "metrics": {
        "moe": 0.02144,
        "pfi": "0.3611",
        "min_val": 0.2941,
        "mean_val": 0.41831
      }
    },
    "virmetrics": {
      "metrics": {
        "moe": 0.02381,
        "pfi": "0.3611",
        "min_val": 0.2,
        "mean_val": 0.30395
      }
    },
    "model_variant": "v4_0_0"
  }
}

Option 2: Prepare forecast payload with segments

Create your forecast payload using the selected segments. Segments use a simplified composite key format.

1
2
3
4
5
6
7
8
9
10
11
{
  "type": "Generate",
  "displayText": "Formula one season testing 2026",
  "additionalDetail": "🚀 Day 1 of 2026 F1 Bahrain testing just dropped BOMBS! Lando Norris edged out Max Verstappen for P1 in McLaren's beast, with Ferrari hot on their heels 🔥. But the real shocker? Cadillac's debutants Sergio Perez & Valtteri Bottas smashed 107 laps – nearly TWO race distances! New kids on the block are HERE to stay. Audi's radical upgrades already turning heads too. Who's your early fave for Australia? 👀 #F12026 #BahrainTesting #LandoNorris",
  "population": {
    "population": "us2",
    "segments": [
      "gender::female"
    ]
  }
}