Setup LLM Provider in VS Code Extensions (Cline, Kilocode)

All models are accessed through Cloudflare AI Gateway (demo-hkmci). Users only need their personal gateway token (cfut_...) — no provider keys required.

👤 User Guide

Step 1 — Get your Gateway Token

API Token: [Ask your admin — starts with cfut_...]

Step 2 — Configure the Extension

Cline:

1. Click the Cline icon in the Activity Bar
2. Click the Settings (gear) icon at the top of the Cline panel
3. Under Provider, select OpenAI Compatible
4. Fill in the fields below

Kilocode:

1. Click the Kilocode icon in the Activity Bar
2. Click your profile/account icon (bottom-left of the Kilocode panel) → Settings
3. Go to the Providers tab
4. Click + Add Provider and select OpenAI Compatible
5. Fill in the fields below

Important — Model field: The model list will not auto-populate. You must type the Model ID directly into the model field (e.g. minimaxai/minimax-m2-maas). Do not wait for a dropdown.

API Key: Enter the cfut_... token as-is. The extension sends it as Authorization: Bearer cfut_... which the gateway accepts.

Step 3 — Select a Model

Use the tables below to pick a Base URL + Model ID pair. Every model uses OpenAI Compatible as the provider type.

Step 4 — Verify it Works

After saving, open a new chat in the extension and send a short message (e.g. “say hi”). If you see a response, the setup is complete. If you get an auth error, double-check the API Key. If you get a model error, make sure you typed the Model ID exactly as shown.

Screenshots

Kilocode:

Cline:

Available Models

Google Gemini (via Google AI Studio)

Base URL:

https://gateway.ai.cloudflare.com/v1/b326904912840c25f63808a1d1e479aa/demo-hkmci/google-ai-studio/v1beta/openai

Prices per million tokens (USD). * = higher price applies when input > 200K tokens.

Model IDs verified working via CF AI Gateway on 2026-04-09.

GCP Vertex AI MaaS — MiniMax, Kimi, GLM

Base URL:

https://gateway.ai.cloudflare.com/v1/b326904912840c25f63808a1d1e479aa/demo-hkmci/google-vertex-ai/v1/projects/mcps-testing-cloudflare-ai/locations/global/endpoints/openapi

Prices per million tokens (USD). Source: GCP Vertex AI partner model pricing.

All model IDs verified working via CF AI Gateway on 2026-04-09.
Newer versions (MiniMax-M2.5, Kimi K2.5, GLM-5.1) exist in GCP Model Garden but are deploy-only — no MaaS endpoint yet.

Cloudflare Workers AI (CF-hosted, no GCP cost)

Base URL:

https://gateway.ai.cloudflare.com/v1/b326904912840c25f63808a1d1e479aa/demo-hkmci/workers-ai/v1

Workers AI uses Cloudflare’s own infrastructure (not GCP). Billed in CF Neurons ($0.011/1K Neurons) with a free daily allowance. These are different model instances from the GCP Vertex AI ones above.

Anthropic Claude (Under Investigation)

Two paths are available — choose based on what the admin has configured:

Option A — Direct Anthropic via Gateway (requires Anthropic BYOK key in gateway) Provider Type: Anthropic Base URL:

https://gateway.ai.cloudflare.com/v1/b326904912840c25f63808a1d1e479aa/demo-hkmci/anthropic

Option B — Claude via GCP Vertex AI (requires Claude access approved in GCP) ⚠️ Pending Provider Type: Anthropic Base URL:

https://gateway.ai.cloudflare.com/v1/b326904912840c25f63808a1d1e479aa/demo-hkmci/google-vertex-ai/v1/projects/mcps-testing-cloudflare-ai/locations/us-east5/publishers/anthropic/models

Prices per million tokens via GCP Vertex AI.

Which Model Should I Use?

🔧 Admin Guide: Gateway Setup

Architecture Overview

Cline / Kilocode
      │  cfut_... (per-user gateway token)
      ▼
Cloudflare AI Gateway (demo-hkmci)
      ├── /google-ai-studio/...  → Google Gemini       [BYOK: AI Studio key]
      ├── /google-vertex-ai/...  → GCP Vertex AI MaaS  [BYOK: service account JSON] ✅
      ├── /workers-ai/v1         → CF Workers AI        [built-in, CF Neurons]
      └── /anthropic             → Anthropic Claude     [BYOK: not configured]

Each user gets their own cfut_... token created from the gateway dashboard. No provider keys are ever exposed to users.

Gateway Details

Provider Keys (BYOK) Configuration

Navigate to: Cloudflare Dashboard → AI → AI Gateway → demo-hkmci → Provider Keys

Always set alias to default so the gateway injects it automatically without requiring users to hold provider keys.

Customer Provisioning

Option A — Dashboard (manual)

AI Gateway → demo-hkmci → Settings → Authenticated Gateway → Create authentication token

• Name: user’s name (e.g. alex)
• Permission: Run

Copy the cfut_... value — it is shown only once.

Option B — Script (automated)

Gateway tokens are standard Cloudflare API tokens and can be created programmatically using /user/tokens.

One-time setup: Create a master CF API token from the dashboard with:

• User → API Tokens → Edit permission

Store it securely — this is the only dashboard step required.

Provision a token for each user:

CF_API_TOKEN=<master-token> \
CF_ACCOUNT_ID=b326904912840c25f63808a1d1e479aa \
python3 /Users/zorro/Dev/cloudflare/create_ai_gateway_token.py \
  --gateway-id demo-hkmci \
  --token-name alex

The script prints the cfut_... value — save it immediately, it will not be shown again.

Script location: /Users/zorro/Dev/cloudflare/create_ai_gateway_token.py

How to Deliver Config to Users

Send users the following (nothing else — no provider keys):

Cloudflare AI Gateway – VS Code Setup

API Token: cfut_...   ← their personal token

For Base URLs, model IDs, and setup steps, see:
https://sharehub.zorro.hk/documents/2026-04-01-vscode-llm-provider-setup.html

Revoke a user

Dashboard: AI Gateway → demo-hkmci → Settings → Authenticated Gateway → [find token] → Delete

API:

# First get the token ID
curl "https://api.cloudflare.com/client/v4/user/tokens" \
  -H "Authorization: Bearer <master-token>"

# Then delete by ID
curl -X DELETE "https://api.cloudflare.com/client/v4/user/tokens/<token-id>" \
  -H "Authorization: Bearer <master-token>"

Revocation is immediate.

GCP Vertex AI Setup

Billing is enabled. Service account key is stored in the gateway as the Google Vertex AI BYOK.

• GCP Project: mcps-testing-cloudflare-ai
• Service Account: cloudflare-ai-gateway-vertex@mcps-testing-cloudflare-ai.iam.gserviceaccount.com
• Active MaaS models: MiniMax-M2, Kimi K2 Thinking, GLM-5, GLM-4.7 (all via locations/global/endpoints/openapi)

Adding a New Provider

1. Get the API key for the provider
2. Go to demo-hkmci → Provider Keys → find provider → click +
3. Paste the key, set alias to default, save
4. Update the User Guide table above with the new Base URL and model IDs
5. Republish this note: /kf-cli:publish 2026-04-01-vscode-llm-provider-setup.md

Captured: 2026-04-01 Last Updated: 2026-04-09 (added Vertex AI MaaS pricing; added Worker proxy for per-customer key provisioning)