Coding agents

Use the same API base URL for every client:

https://api.llmfaucet.dev/v1

Use free for anonymous testing or a personal llmfaucet_... key from the dashboard. Use auto:coding when the request is code-focused.

Cline and Roo Code

Provider: OpenAI Compatible
Base URL: https://api.llmfaucet.dev/v1
API key: free or llmfaucet_...
Model: auto:coding

Keep the base URL ending in /v1. Do not paste a key into a public issue, screen recording, or repository.

Aider

aider \
  --openai-api-base https://api.llmfaucet.dev/v1 \
  --openai-api-key free \
  --model openai/auto:coding

For a personal key, set --openai-api-key from an environment variable rather than placing the value in shell history.

Claude Code

The Worker exposes the Anthropic-compatible messages surface at the API root:

export ANTHROPIC_BASE_URL="https://api.llmfaucet.dev"
export ANTHROPIC_AUTH_TOKEN="free"
claude

Use ANTHROPIC_AUTH_TOKEN; do not configure ANTHROPIC_API_KEY for this integration.

Codex CLI

[model_providers.llmfaucet]
name = "llmfaucet"
base_url = "https://api.llmfaucet.dev/v1"
env_key = "LLMFAUCET_API_KEY"
wire_api = "responses"
query_params = {}

export LLMFAUCET_API_KEY=free

The Responses surface is a compatibility shim. Check the preview status page before using it for a long-running workflow.

Continue

models:
  - name: llmfaucet-coding
    provider: openai
    model: auto:coding
    apiBase: https://api.llmfaucet.dev/v1
    apiKey: free

Compatibility notes

• Prefer chat completions for general coding agents.
• Streaming is SSE and should be handled incrementally.
• A route can fail over only to a compatible capability class.
• Read X-Routed-Via, X-Fallback-Attempts, and rate-limit headers when diagnosing a client issue.
• Provider names and available models are not fixed; use GET /v1/models.

See limits and early testers before connecting an automated coding loop.