Clef · Try the live demo →
Heabsy

Kilo Code custom provider: connect Heabsy

Kilo Code takes any OpenAI-compatible API as a custom provider. In VS Code open Settings → Providers → Custom provider, choose the API type "OpenAI Compatible" — not "OpenAI Responses" — set the base URL https://api.heabsy.com/v1, paste the key and fetch the model list. For the CLI, the same provider goes into kilo.jsonc.

~/.config/kilo/kilo.jsonc
{
  "provider": {
    "heabsy": {
      "options": {
        "baseURL": "https://api.heabsy.com/v1",
        "apiKey": "{env:HEABSY_API_KEY}"
      },
      "models": {
        "qwen38": { "tool_call": true, "reasoning": true, "limit": { "context": 262144, "output": 32768 } }
      }
    }
  }
}
Create a Heabsy key OpenAI base URL https://api.heabsy.com/v1 · Anthropic base URL https://api.heabsy.com

In VS Code

  1. Settings → Providers → Custom provider.
  2. Provider ID heabsy, display name Heabsy.
  3. Provider API: OpenAI Compatible. Heabsy does not serve the Responses API, so the "OpenAI Responses" option ends in a 404.
  4. Base URL https://api.heabsy.com/v1, API key, then Fetch models — Kilo reads /v1/models and lists exactly the models your key may call.
  5. Set context and output limits for the models you use.

Why the limits matter

Kilo Code turns automatic context management off for models without limit.context and limit.output. Take the numbers from the table below. The {env:…} reference works only in the trusted global config, not in a project's kilo.json.

Which model should I pick?

Kilo's agent modes need tool calling; every model in the table has it. Our own models run in the EEA without storing prompts.

ModelIDContextMax outputInput / 1MOutput / 1M
Qwen3.8 27B Fableqwen38-fable512K32K$0.095$0.58
Qwen3.8 27B Queenqwen38-queen512K32K$0.095$0.58
GLM 5.3 Flash Uncensoredglm53-flash-unc1M128K$0.1739$0.6957
MiMo V2.6 Flash Uncensoredmimo-v26-flash-unc1M64K$0.4404$1.4285
MiMo V2.6 Flashmimo-v26-flash1M64K$0.4762$1.4286
MiMo V2.6 Flash Abliteratedmimo-v26-flash-abl1M64K$0.4762$1.4286
GLM 5.3glm531M128K$0.5217$0.9565
Qwen3.8 27B Uncensored Fast by Heabsyqwen38-unc-fast512K230K$0.7904$2.4952
Qwen3.8 27B Uncensoredqwen38-unc1M230K$0.8095$2.5238
Qwen3.8 27Bqwen38256K32K$1.00$3.00
GLM 5.3 Uncensoredglm-5.3-uncensored1M128K$1.1905$2.1429
CyberHeabsycyberheabsy1M64K$1.4286$4.2857

All models above run on Heabsy hardware in the EEA; prompts and completions are not stored. Thecatalog also lists 23 partner-hosted models — Claude, GPT-6, Kimi K3, DeepSeek V4 and others — with the same key and invoice, but without zero data retention and without an EEA guarantee.

Verify it works

Pick heabsy/qwen38, switch to Code mode and ask Kilo to read a file. A correct answer means the connection and tool calling work; the request appears in the Logs page of your console.

If the tool shows an error you cannot read, test the key on its own first — one short request outside the tool:

curl
curl https://api.heabsy.com/v1/chat/completions \
  -H "Authorization: Bearer $HEABSY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen38","max_tokens":16,"messages":[{"role":"user","content":"Say hi"}]}'

Troubleshooting

What you seeWhat to change
404 on every requestThe provider API is set to "OpenAI Responses". Switch to "OpenAI Compatible".
Fetch models returns nothingCheck the key, then the base URL — it must end in /v1.
Long sessions fail instead of condensingAdd limit.context and limit.output for the model.
403 with an HTML page ("Sorry, you have been blocked")No key was sent at all: the key variable is empty or not set where the tool runs. Requests without a key are stopped before they reach the API; a wrong key gets a JSON 401 instead.

Where do my prompts go?

From Kilo Code straight to api.heabsy.com. On our own models the request is processed in the European Economic Area and the text is not stored — we keep what billing needs: the model, the token counts and the time. Partner-hosted models are processed by third-party inference providers; each is marked in the catalog, and none of them carries the no-retention promise. The terms are in the data processing agreement, the invoice comes from an EU company.

FAQ

Coming from Roo Code?+

Kilo Code reads Roo's .roomodes on first start, so custom modes carry over. Add Heabsy as a custom provider and your Roo workflow continues on our models.

Which model should I start with?+

qwen38 for most coding work; glm53 when you need a 1,048,576-token window at $0.52 / $0.96 per million input / output tokens.

Can I use Kilo's own gateway and Heabsy side by side?+

Yes. Providers are separate; switch the model in the picker.

Create a key and send the first request All integrations

Also set up:Cline · OpenCode · Claude Code

Kilo Code is a trademark of its owners; Heabsy is not affiliated with it.