Connect → Kilo Code

Use AxForge with Kilo Code

Kilo Code is a coding agent for VS Code. Add AxForge as a Custom provider in its settings: Kilo reads our model list, you tick the models, and they show in its model picker.

Set it up

1. Open Providers

   Open Kilo Code from its icon in the side bar (the first time it asks how you want to work: either choice is fine). Click the gear above its panel, then Providers. At Custom provider, press Connect.

   

2. Describe AxForge

   Provider ID axforge, Display name AxForge, Provider API OpenAI Compatible, Base URL https://api.axforge.ai/v1, and your key in API key (it shows as dots).

   

3. Pick the models from our list

   With the address and key in, Kilo reads our model list and ticks all of it. Press Deselect all, then tick qwen3.8-27b-nvfp4, glm-5.3-flash and mimo-v2.6-flash (type a name in the search box to find it) and press Add 3 model(s). The rest of the list are short aliases and our image, speech, embedding and OCR models, which a coding agent does not use.

   

4. Submit

   The three models are now rows of the provider. Press Submit.

   

5. Tell Kilo the models' size

   The form has no field for it. In Kilo's settings press Global Config (top right): it opens ~/.config/kilo/kilo.jsonc, where Submit saved the provider (the key is kept apart). Give each model a limit and save:

   "models": {
     "glm-5.3-flash": { "name": "glm-5.3-flash", "limit": { "context": 262144, "output": 16384 } },
     "mimo-v2.6-flash": { "name": "mimo-v2.6-flash", "limit": { "context": 262144, "output": 16384 } },
     "qwen3.8-27b-nvfp4": { "name": "qwen3.8-27b-nvfp4", "limit": { "context": 262144, "output": 16384 } }
   }

   Our models read 262,144 tokens at once. With the limit, Kilo shows it beside each chat and shortens a long session before it reaches that.

6. Pick the model in the chat

   Click the model name under the message box, type qwen3.8-27b-nvfp4 and press Select.

   

7. Give it a task

   Ask in plain words. Kilo reads the file, edits it and says what it did.

   

The settings

Which model for what: qwen3.8-27b-nvfp4 is the everyday default and reads images, glm-5.3-flash thinks before it answers, mimo-v2.6-flash is the fastest.

Check that it works

Under the message box it says qwen3.8-27b-nvfp4, and a task gets done. The calls show in the console under API Keys → Recent calls within a minute.

When it does not work

Test a call says in one click whether a problem is your key, your balance or us.