Tool Use Now Available on Gonka
The recent v0.2.10 network upgrade brought tool calling support to Gonka’s main model. The upgrade enables tool use arguments (--enable-auto-tool-choice, --tool-call-parser hermes) for Qwen/Qwen3-235B-A22B-Instruct-2507-FP8 with a validation threshold of 0.958.
Access Gonka AI Through Mingles Gateway
You can now connect OpenClaw to Gonka AI’s decentralized compute through the public gateway at https://gonka-gateway.mingles.ai/
Getting Started:
- Free trial: you get 0.1 GNK credit to get started
- OpenAI-compatible API endpoint
- Powered by Gonka’s distributed GPU network
Don’t have a Gonka account yet?
You can create one through Keplr wallet (visit keplr.app and add the Gonka network) or generate one using the inferenced CLI. For detailed setup instructions, check the Gonka documentation.
Step 1: Get Your API Key
You have two options for obtaining an API key:
Option A: Web Dashboard (Easiest)
- Visit https://gonka-gateway.mingles.ai/
- Connect with Keplr wallet (Chrome recommended)
- Navigate to API Keys menu
- Click Create API Key
- Copy the key (starts with
sk-)
Security Note: Don’t import your cold account seed phrase into Keplr. Instead, create a new Gonka account in Keplr specifically for web access. You can top it up from your main account if needed.
Option B: Generate from Cold Account Machine
If you want to use your existing cold account without importing it to Keplr:
Visit the agent-friendly setup guide: Agent / Bot integration — Gonka Gateway
This provides instructions for generating an API key directly from your cold account machine using the command line.
Step 2: Configure OpenClaw
Run one of these commands:
Fresh setup:
openclaw reset
openclaw onboard
Existing setup:
openclaw configure
During configuration:
- When prompted for model provider, select: Custom gateway
- Select: OpenAI compatible
- Enter details:
- URL:
https://gonka-gateway.mingles.ai/v1 - API Key: Your key from Step 1 (starts with
sk-)
- URL:
Step 3: Fix Context Window (If Needed)
If you encounter context window limitations, update your configuration:
nano ~/.openclaw/openclaw.json
Find and update these values:
"contextWindow": 200000,
"maxTokens": 8192
Save and restart the gateway:
openclaw gateway stop && openclaw gateway start