Integration guide

Get your API key from your project dashboard, then swap your existing base URL and key. No SDK changes needed — TokenFund speaks OpenAI's protocol.

Roo Code supports custom OpenAI-compatible providers. Set it up once per project or globally.

Base URL:  https://tokenfund.dev/api/v1
API Key:   tf_proj_your_key_here
Model:     claude-sonnet-4-5

Cline (formerly Claude Dev) supports custom OpenAI-compatible endpoints.

OpenAI API Key:     tf_proj_your_key_here
OpenAI Base URL:    https://tokenfund.dev/api/v1
Model ID:           claude-sonnet-4-5

Cursor supports custom models via its OpenAI-compatible settings.

API Key:   tf_proj_your_key_here
Base URL:  https://tokenfund.dev/api/v1

TokenFund is a drop-in replacement for the OpenAI SDK base URL. No other changes needed.

curl https://tokenfund.dev/api/v1/chat/completions \
  -H "Authorization: Bearer tf_proj_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-5",
    "messages": [{"role": "user", "content": "Hello!"}],
    "max_tokens": 256
  }'

Use the Proxy API to forward requests to provider endpoints (models list, search, generation stats, etc.). For chat completions and messages with wallet billing, use /api/v1/chat/completions. The proxy forwards everything else. Streaming works where the upstream supports it.

curl https://tokenfund.dev/api/proxy/anthropic/v1/messages \
  -H "Authorization: Bearer tf_proj_your_key" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-sonnet-4-5","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'

curl "https://tokenfund.dev/api/proxy/openrouter/api/v1/models" \
  -H "Authorization: Bearer tf_proj_your_key"

curl https://tokenfund.dev/api/proxy/exa/search \
  -H "Authorization: Bearer tf_proj_your_key" \
  -H "Content-Type: application/json" \
  -d '{"query":"machine learning","numResults":5}'

Set the provider's API key in your environment (e.g. ANTHROPIC_API_KEY, OPENROUTER_API_KEY, EXA_API_KEY, FAL_KEY) for platform billing. For hackathons, organizer keys are used when configured.

Pass the model name in the model field. Anthropic models are converted from OpenAI format automatically.