Docs

Integrations

SOKKAN Router speaks the OpenAI Chat Completions API, so any agent harness that accepts a custom OpenAI-compatible provider can use it. Below, the settings for the most used tools. Each block links to the tool's own documentation, which is the reference if a menu or a key name changes in a newer release.

Settings to use

Base URLhttps://router.sokkan.ch/v1
API keyskr_… (create one in your account)
Model idqwen/qwen3.8-27b, or any id from GET /v1/models

Put the key in an environment variable rather than in a file you commit. The examples use SOKKAN_API_KEY. See Quickstart for authentication, streaming, errors and limits. Current model ids:

curl -s https://router.sokkan.ch/v1/models

Which model

All figures below come from Models & pricing and GET /v1/models.

UseModel idWhy
Agents and codeqwen/qwen3.8-27b131,072-token context, up to 32,768 output tokens, zone EU.
Simple, high-volume tasksopenai/gpt-oss-120bLowest per-token prices in the catalogue ($0.10 in / $0.50 out per million tokens), 131,072-token context, zone EU.
Requests that must stay in Switzerlandnone on sale right nowNo model is currently on sale in zone CH. A request sent with X-Data-Location: CH is refused (no_provider), never served in another zone; Models & pricing shows the zone of every model.

A coding agent sends long prompts, so check the context length of the model you pick against what the tool needs. The tools below ask for a context window or a max output size in their settings: use the figures above.

Zone filter and region header

Every response carries x-sokkan-region (CH, EU or US): the zone that generated it. To restrict where a request may run, send the request header X-Data-Location: CH (or EU), or "provider": {"data_location": "CH"} in the body. If the model does not run there, you get 400 no_provider and nothing is sent elsewhere. Details in Data location.

Not every tool lets you add a request header. Where it does (Kilo Code has a Headers field, the OpenAI SDKs have default headers), add X-Data-Location. Where it does not, pick a model that runs in the zone you need (see the table above): the model id decides the zone.

Hermes Agent (Nous Research)

Docs: Hermes Agent, providers. Hermes reads ~/.hermes/config.yaml; the key can come from an environment variable named by key_env, set in ~/.hermes/.env. Or run hermes model and choose Custom endpoint.

# ~/.hermes/config.yaml
model:
  default: qwen/qwen3.8-27b
  provider: custom
  base_url: https://router.sokkan.ch/v1
  key_env: SOKKAN_API_KEY
  context_length: 131072   # only if auto-detection fails

# ~/.hermes/.env
SOKKAN_API_KEY=skr_…

OpenClaw

Docs: OpenClaw, custom providers. Declare the provider under models.providers in your OpenClaw config, with the openai-completions API type, then select it as provider-id/model-id.

// OpenClaw config (JSON5)
{
  agents: { defaults: { model: { primary: "sokkan/qwen/qwen3.8-27b" } } },
  models: {
    providers: {
      sokkan: {
        baseUrl: "https://router.sokkan.ch/v1",
        apiKey: "${SOKKAN_API_KEY}",
        api: "openai-completions",
        models: [
          {
            id: "qwen/qwen3.8-27b",
            name: "Qwen3.8-27B",
            reasoning: false,
            input: ["text"],
            cost: { input: 0.20, output: 2.50, cacheRead: 0.05, cacheWrite: 0 },
            contextWindow: 131072,
            maxTokens: 32768
          }
        ]
      }
    }
  }
}

Cline

Docs: Cline, OpenAI Compatible. Open Cline's settings and fill in:

API ProviderOpenAI Compatible
Base URLhttps://router.sokkan.ch/v1
API Keyskr_…
Modelqwen/qwen3.8-27b

Under Model Configuration, set Context Window to 131072 and Max Output Tokens to your choice (up to 32768 for this model).

Kilo Code

Docs: Kilo Code, OpenAI Compatible. In Settings, open the Providers tab, scroll down and click Custom provider:

Provider IDsokkan
Display nameSOKKAN Router
Provider APIOpenAI Compatible
Base URLhttps://router.sokkan.ch/v1
API keyskr_…
Modelsauto-fetched from /v1/models, or add qwen/qwen3.8-27b by hand
Headersoptional: X-Data-Location = CH

Roo Code

Docs: Roo Code, OpenAI Compatible. In Roo Code's settings:

API ProviderOpenAI Compatible
Base URLhttps://router.sokkan.ch/v1
API Keyskr_…
Modelqwen/qwen3.8-27b

Note. Roo Code's documentation says it uses native tool calling only, so the model you pick must support it. Test the model with a small task before a long session.

Continue

Docs: Continue, OpenAI provider. Add a model with the openai provider and your own apiBase in config.yaml:

# config.yaml
name: My Config
version: 0.0.1
schema: v1

models:
  - name: SOKKAN Router, Qwen3.8-27B
    provider: openai
    model: qwen/qwen3.8-27b
    apiBase: https://router.sokkan.ch/v1
    apiKey: skr_…

Aider

Docs: Aider, OpenAI compatible APIs. Set the two environment variables and prefix the model id with openai/:

export OPENAI_API_BASE=https://router.sokkan.ch/v1
export OPENAI_API_KEY=$SOKKAN_API_KEY
aider --model openai/qwen/qwen3.8-27b

LiteLLM

Docs: LiteLLM, OpenAI-compatible endpoints. Prefix the model with openai/ and pass api_base. The base URL must end in /v1.

# pip install litellm
import os, litellm

response = litellm.completion(
    model="openai/qwen/qwen3.8-27b",
    api_key=os.environ["SOKKAN_API_KEY"],
    api_base="https://router.sokkan.ch/v1",
    messages=[{"role": "user", "content": "Hello"}],
)

OpenAI SDKs (Python, JavaScript)

Docs: openai-python, openai-node. Change the base URL and the key; the rest of your code stays the same. Python also reads the OPENAI_BASE_URL and OPENAI_API_KEY environment variables.

# pip install openai
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://router.sokkan.ch/v1",
    api_key=os.environ["SOKKAN_API_KEY"],
    default_headers={"X-Data-Location": "EU"},   # optional zone filter
)
resp = client.chat.completions.create(
    model="qwen/qwen3.8-27b",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

curl

curl -i https://router.sokkan.ch/v1/chat/completions \
  -H "Authorization: Bearer $SOKKAN_API_KEY" \
  -H "X-Data-Location: EU" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3.8-27b",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

The x-sokkan-region response header (visible with -i) tells you where the answer was generated.

Another tool?

Any client that accepts an OpenAI base URL works the same way. If one does not, write to [email protected] and quote the x-request-id of a failed request.