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

# Kilo Code setup guide

> Connect Kilo Code to the powertokens platform through OpenAI Compatible or Anthropic provider settings.

Kilo Code can connect to the `powertokens platform` through two provider paths:

* OpenAI Compatible provider for text models that work through `/v1/chat/completions`.
* Anthropic provider with a custom base URL for Anthropic-path models validated as default tool-workflow choices.

Choose the path by protocol support. OpenAI-compatible support does not mean the same model can also be used through the Anthropic provider.

If you need to choose the endpoint first, see [Text model protocols and endpoints](/en/ecosystem-tools/text-model-protocols).

## Choose a path

| Goal                                   | Kilo Code provider                       | Base URL                        | Model scope                                                                                             |
| -------------------------------------- | ---------------------------------------- | ------------------------------- | ------------------------------------------------------------------------------------------------------- |
| Use OpenAI-compatible text models      | OpenAI Compatible or OpenAI (Compatible) | `https://api.powertokens.ai/v1` | Text models that support `/v1/chat/completions`                                                         |
| Use Anthropic-path models              | Anthropic with custom base URL           | `https://api.powertokens.ai/v1` | Recommended Zhipu, MiniMax, Qwen, BytePlus DeepSeek, and re-validated Seed models that support Messages |
| Use OpenAI-compatible preferred models | OpenAI Compatible                        | `https://api.powertokens.ai/v1` | Text models whose upstream provider does not support `/v1/messages`, such as `seed-1-6-flash-250715`    |

## OpenAI Compatible setup

Create a provider profile in Kilo Code:

| Field                 | Value                                                                        |
| --------------------- | ---------------------------------------------------------------------------- |
| **API Provider**      | **OpenAI Compatible** or **OpenAI (Compatible)**                             |
| **Base URL**          | `https://api.powertokens.ai/v1`                                              |
| **API Key**           | Your `powertokens platform` API key                                          |
| **Model**             | For example `glm-5.2`, `MiniMax-M3`, or `seed-2-0-pro-260328`                |
| **Max output tokens** | Set for your task and selected model                                         |
| **Context window**    | Set conservatively if you are unsure                                         |
| **Image support**     | Enable only when the model and the platform API publicly support image input |

Verify the same path with:

```bash theme={null}
curl -X POST "https://api.powertokens.ai/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_POWERTOKENS_API_KEY" \
  --data '{
    "model": "glm-5.2",
    "messages": [
      {"role": "user", "content": "Reply with ok."}
    ],
    "stream": false
  }'
```

## Anthropic provider setup

Use Kilo Code's Anthropic provider when you want the Claude/Anthropic path:

| Field                   | Value                                  |
| ----------------------- | -------------------------------------- |
| **API Provider**        | **Anthropic**                          |
| **Use custom base URL** | Enabled                                |
| **Custom base URL**     | `https://api.powertokens.ai/v1`        |
| **Anthropic API Key**   | Your `powertokens platform` API key    |
| **Model**               | One of the recommended model IDs below |

Kilo Code's Anthropic provider appends `/messages` to **Custom base URL**. Set the field to the version path `https://api.powertokens.ai/v1` so the final request reaches the platform's `/v1/messages` endpoint.

Recommended Anthropic tool-workflow models:

| Upstream provider | Models                                                                                                      |
| ----------------- | ----------------------------------------------------------------------------------------------------------- |
| Zhipu             | `glm-4.5-air`, `glm-4.7`, `glm-4.7-flash`, `glm-5`, `glm-5-turbo`, `glm-5.2`                                |
| MiniMax           | `MiniMax-M2.5`, `MiniMax-M2.5-highspeed`, `MiniMax-M2.7`, `MiniMax-M2.7-highspeed`, `MiniMax-M3`            |
| Alibaba Qwen      | `qwen3-coder-plus`, `qwen3-max`, `qwen3.5-flash`, `qwen3.6-plus`                                            |
| BytePlus DeepSeek | `deepseek-v3-2-251201`                                                                                      |
| BytePlus Seed     | `seed-1-6-250915`, `seed-1-8-251228`, `seed-2-0-lite-260228`, `seed-2-0-mini-260215`, `seed-2-0-pro-260328` |

When you verify the Anthropic path with curl, use the full endpoint path:

```bash theme={null}
curl -X POST "https://api.powertokens.ai/v1/messages" \
  -H "Content-Type: application/json" \
  -H "x-api-key: YOUR_POWERTOKENS_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  --data '{
    "model": "glm-5.2",
    "max_tokens": 64,
    "messages": [
      {"role": "user", "content": "Reply with ok."}
    ],
    "stream": false
  }'
```

## Models not recommended for Anthropic provider

Use these models through OpenAI Compatible, not through the Anthropic provider, even if model discovery declares `anthropic`:

| Upstream provider | Models                  |
| ----------------- | ----------------------- |
| BytePlus          | `seed-1-6-flash-250715` |

## Validation checklist

1. Send a tiny prompt that only asks for `ok`.
2. Run one real code-edit task and confirm Kilo Code can read files, propose edits, and finish.
3. If you enabled streaming, validate a long response.
4. Validate tool calling, image input, and caching separately before relying on them.

## Troubleshooting

| Symptom                                         | Action                                                                                                     |
| ----------------------------------------------- | ---------------------------------------------------------------------------------------------------------- |
| `Invalid API Key`                               | Check the key and provider path                                                                            |
| `Model Not Found`                               | Check the exact model ID; enter `glm-5.2` manually if discovery does not list it                           |
| Anthropic provider tool workflow fails          | The model is not currently suitable for Kilo Code Anthropic-provider workflows; use OpenAI Compatible      |
| Anthropic provider returns an empty response    | Do not use that model as the Anthropic-provider default until you validate it; use OpenAI Compatible first |
| Tool output or fields differ from upstream docs | Check whether the field is publicly supported by the platform                                              |