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

# Create LLM Integration

> Creates a new LLM integration for a runner.

`Unary` · [`Runner Configurations`](https://ona.com/docs/api-reference/generated/runner-configuration/overview)

Creates a new LLM integration for a runner.

Use this method to:

* Configure LLM access
* Set up LLM integrations
* Enable LLM usage

## Endpoint

```text theme={null}
POST /api/gitpod.v1.RunnerConfigurationService/CreateLLMIntegration
```

Send a Bearer token as described in [Authentication](https://ona.com/docs/api-reference.md#authenticate-requests). If your organization uses a custom management-plane domain, replace `https://app.ona.com` with that domain.

## Request example

```bash cURL theme={null}
export ONA_HOST=https://app.ona.com
export ONA_API_KEY=<your-token>

curl --request POST \
  --url "$ONA_HOST/api/gitpod.v1.RunnerConfigurationService/CreateLLMIntegration" \
  --header "Authorization: Bearer $ONA_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
  "runnerId": "<runner-id>"
}'
```

```python Python theme={null}
import gitpod.v1.runner_configuration_pb2 as runner_configuration_pb2
from ona_sdk import create_client_from_env

ona = create_client_from_env()
request = runner_configuration_pb2.CreateLLMIntegrationRequest(
    runner_id="<runner-id>",
)
response = ona.services.runner_configuration.create_llm_integration(request)
print(response)
```

```typescript TypeScript theme={null}
import { create } from "@bufbuild/protobuf";
import { createClientFromEnv } from "@gitpod/sdk";
import { CreateLLMIntegrationRequestSchema } from "@gitpod/sdk/gitpod/v1/runner_configuration_pb";

async function main() {
  const ona = createClientFromEnv();
  const request = create(CreateLLMIntegrationRequestSchema, {
    runnerId: "<runner-id>",
  });
  const response = await ona.services.runnerConfiguration.createLLMIntegration(request);
  console.log(response);
}

main().catch(console.error);
```

```go Go theme={null}
package main

import (
	"context"
	"fmt"
	"log"

	"connectrpc.com/connect"
	"github.com/gitpod-io/gitpod-sdk-go/sdk"
	gitpodpb "github.com/gitpod-io/gitpod-sdk-go/v1"
)

func main() {
	ona, err := sdk.NewFromEnv()
	if err != nil {
		log.Fatal(err)
	}

	request := connect.NewRequest(&gitpodpb.CreateLLMIntegrationRequest{
		RunnerId: "<runner-id>",
	})
	response, err := ona.Services.RunnerConfiguration.CreateLLMIntegration(context.Background(), request)
	if err != nil {
		log.Fatal(err)
	}
	fmt.Println(response.Msg)
}
```

```json Request body theme={null}
{
  "runnerId": "<runner-id>"
}
```

## Request

`gitpod.v1.CreateLLMIntegrationRequest`

| Field | Type | Required | Description |
| - | - | - | - |
| `runnerId` | string | No | Constraints: `string.uuid=true`. |
| `models` | array of [SupportedModel](#enum-gitpod-v1-supported-model) | No | models represents the LLM models this integration supports Constraints: `repeated.items.enum.defined_only=true, repeated.min_items=1`. |
| `endpoint` | string | No | |
| `apiKey` | string | No | api\_key is the LLM provider's API key. |
| `maxTokens` | 64-bit integer string | No | max\_tokens is the maximum number of tokens to generate before stopping. If set to 0, the model's default will be used. |

## Response

`gitpod.v1.CreateLLMIntegrationResponse`

| Field | Type | Required | Description |
| - | - | - | - |
| `id` | string | No | id is a uniquely generated identifier for the LLM integration Constraints: `string.uuid=true`. |

## Related types

> [accordion] SupportedModel
>
> SupportedModel enumerates the LLM models available for agent executions
>
> | Value | Number | Description |
> | - | -: | - |
> | `SUPPORTED_MODEL_UNSPECIFIED` | 0 | |
> | `SUPPORTED_MODEL_SONNET_3_5` | 1 | |
> | `SUPPORTED_MODEL_SONNET_3_7` | 2 | |
> | `SUPPORTED_MODEL_SONNET_3_7_EXTENDED` | 3 | |
> | `SUPPORTED_MODEL_SONNET_4` | 4 | |
> | `SUPPORTED_MODEL_SONNET_4_EXTENDED` | 5 | |
> | `SUPPORTED_MODEL_SONNET_4_5` | 8 | |
> | `SUPPORTED_MODEL_SONNET_4_5_EXTENDED` | 9 | |
> | `SUPPORTED_MODEL_SONNET_4_6` | 18 | |
> | `SUPPORTED_MODEL_SONNET_4_6_EXTENDED` | 19 | |
> | `SUPPORTED_MODEL_SONNET_5` | 32 | |
> | `SUPPORTED_MODEL_OPUS_4` | 6 | |
> | `SUPPORTED_MODEL_OPUS_4_EXTENDED` | 7 | |
> | `SUPPORTED_MODEL_OPUS_4_5` | 14 | |
> | `SUPPORTED_MODEL_OPUS_4_5_EXTENDED` | 15 | |
> | `SUPPORTED_MODEL_OPUS_4_6` | 16 | |
> | `SUPPORTED_MODEL_OPUS_4_6_EXTENDED` | 17 | |
> | `SUPPORTED_MODEL_OPUS_4_7` | 22 | |
> | `SUPPORTED_MODEL_OPUS_4_8` | 31 | |
> | `SUPPORTED_MODEL_HAIKU_4_5` | 21 | |
> | `SUPPORTED_MODEL_OPENAI_4O` | 10 | |
> | `SUPPORTED_MODEL_OPENAI_4O_MINI` | 11 | |
> | `SUPPORTED_MODEL_OPENAI_O1` | 12 | |
> | `SUPPORTED_MODEL_OPENAI_O1_MINI` | 13 | |
> | `SUPPORTED_MODEL_OPENAI_AUTO` | 23 | SUPPORTED\_MODEL\_OPENAI\_AUTO flags a request as OpenAI-bound without encoding a specific model slug. The actual model is chosen by the client (today: native Codex via \~/.codex/config.toml) and captured from the upstream Responses-API response.model field for metering and rate-card lookup. This keeps the proto stable across OpenAI model-catalog churn. Reserved numbers 24-30 are intentionally left free for future OpenAI routing sentinels if we ever need to distinguish sub-families. |


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