Setup guides

Every guide is the same two changes: the base URL and the key. Model ids come straight from the catalog.

Anthropic format: https://api.halfbill.uk
OpenAI format: https://api.halfbill.uk/v1
Key: dashboard → API key

Claude Code

Two environment variables. Switch models with /model.

Anthropic API format

01Copy your Halfbill key from the dashboard.
02Export ANTHROPIC_BASE_URL=https://api.halfbill.uk and ANTHROPIC_AUTH_TOKEN=<key> in your shell profile.
03Optionally pin a default with ANTHROPIC_MODEL=claude-sonnet-5; ANTHROPIC_SMALL_FAST_MODEL=claude-haiku-4-5 keeps the helper model cheap.
04Run claude. Prompt caching keeps working — Claude Code manages its own cache markers.

bash

export ANTHROPIC_BASE_URL=https://api.halfbill.uk
export ANTHROPIC_AUTH_TOKEN=<your Halfbill key>
export ANTHROPIC_MODEL=claude-sonnet-5
claude

Cursor

Override the OpenAI base URL in Cursor's model settings.

OpenAI API format

01Open Cursor Settings → Models.
02Paste your Halfbill key in the OpenAI API key field.
03Enable “Override OpenAI base URL” and set it to https://api.halfbill.uk/v1.
04Add a custom model named claude-sonnet-5 or gpt-6-astra, then pick it in Chat or Agent mode. Tab autocomplete stays on Cursor's own backend.

bash

cursor .

Codex CLI

Add a provider block to ~/.codex/config.toml.

OpenAI API format

01Export HALFBILL_API_KEY=<key> in your shell profile.
02Add the provider block above to ~/.codex/config.toml.
03Run codex. Use /model to switch between gpt-6-astra, gpt-5.5 and the Claude models.

bash

export HALFBILL_API_KEY=<your Halfbill key>
codex

~/.codex/config.toml

model = "gpt-6-astra"
model_provider = "halfbill"

[model_providers.halfbill]
name = "Halfbill"
base_url = "https://api.halfbill.uk/v1"
env_key = "HALFBILL_API_KEY"
wire_api = "responses"

Cline

Pick the OpenAI Compatible provider in Cline's settings.

OpenAI API format

01Open the Cline settings panel in VS Code.
02Set API Provider to “OpenAI Compatible”.
03Base URL: https://api.halfbill.uk/v1 · API key: your Halfbill key · Model ID: claude-sonnet-5.
04Enable prompt caching in Cline's provider settings if it is offered — Claude models support it.

bash

cline

Continue

One model block in config.yaml.

OpenAI API format

01Open ~/.continue/config.yaml (or the Continue settings UI).
02Add the model block above with your Halfbill key.
03Duplicate the block for any other catalog model you want in the picker.

~/.continue/config.yaml

models:
  - name: Claude Sonnet 5 (Halfbill)
    provider: openai
    model: claude-sonnet-5
    apiBase: https://api.halfbill.uk/v1
    apiKey: <your Halfbill key>
    roles: [chat, edit, apply]

OpenCode

Add Halfbill as a custom provider in opencode.json.

OpenAI API format

01Export HALFBILL_API_KEY=<key> in your shell profile.
02Add the provider block above to your OpenCode config.
03Run opencode and pick a Halfbill model with /models.

bash

export HALFBILL_API_KEY=<your Halfbill key>
opencode

~/.config/opencode/opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "halfbill": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Halfbill",
      "options": {
        "baseURL": "https://api.halfbill.uk/v1",
        "apiKey": "{env:HALFBILL_API_KEY}"
      },
      "models": {
        "claude-sonnet-5": {
          "name": "Claude Sonnet 5"
        },
        "gpt-6-astra": {
          "name": "GPT-6 Astra"
        }
      }
    }
  }
}

OpenAI & Anthropic SDKs

Any SDK: change the base URL and the key.

OpenAI API format

01OpenAI-format clients: base URL https://api.halfbill.uk/v1 (/chat/completions and /responses).
02Anthropic-format clients: base URL https://api.halfbill.uk (/v1/messages).
03Use catalog model ids as-is; streaming, tool calls, vision and structured outputs pass through unchanged.

bash

export OPENAI_BASE_URL=https://api.halfbill.uk/v1
export OPENAI_API_KEY=<your Halfbill key>
curl https://api.halfbill.uk/v1/chat/completions -H "Authorization: Bearer $OPENAI_API_KEY" -d '{"model":"gpt-5.5","messages":[{"role":"user","content":"hi"}]}'

Need the full API reference (endpoints, rate limits, caching)? Halfbill Docs