APIreference

1. 01

   Get credits

   Buy credits with USDG or deposit USDG in the Drip app. Requests are paid from this balance.

2. 02

   Create an API key

   Open the console and create a key. It is shown once, so store it right away, for example as DRIP_API_KEY.

3. 03

   Send a request

   Use the base URL below with any OpenAI SDK, or plain HTTP.

curl https://usedrip.lat/v1/chat/completions \
  -H "Authorization: Bearer $DRIP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-haiku-5.5",
    "messages": [{"role": "user", "content": "Say hello in five words."}]
  }'

anthropic/claude-haiku-5.5 is the model ID for Claude Haiku 5.5, the one model Drip serves. See Models for its prices and limits.

Authorization: Bearer drip_sk_…

• Keys start with drip_sk_, followed by 43 random characters.
• Drip stores only a SHA-256 hash of the key and its last four characters. A lost key cannot be shown again; rotate it instead.
• Each key belongs to one Drip account, and all its requests are paid from that account’s credits.
• The API sends CORS headers, so a browser can call it. Still, keep keys on your server: anyone who can read a key can spend your credits.

Actions

• The console also has usage by key, model and day.

# No API key needed
curl https://usedrip.lat/v1/models

# One model
curl https://usedrip.lat/v1/models/anthropic/claude-haiku-5.5

Each model object has OpenAI’s shape plus a drip field with prices, limits, capabilities and fallbacks.

{
  "id": "anthropic/claude-haiku-5.5",
  "object": "model",
  "created": 1760000000,
  "owned_by": "anthropic",
  "drip": {
    "name": "Claude Haiku 5.5",
    "provider": "Anthropic",
    "pricing": {
      "currency": "USD",
      "unit": "1M tokens",
      "input": 0.105,
      "output": 0.525,
      "includes_service_margin": 0.05,
      "placeholder": false
    },
    "context_window": 64000,
    "max_output_tokens": 32000,
    "capabilities": ["text", "vision", "files", "tools", "json", "reasoning"],
    "good_for": "Fast, capable answers, writing, coding help, documents and images.",
    "fallbacks": []
  }
}

Current model

Prices are in USD per 1M tokens and already include Drip’s 5% service margin.

Request body

Fields not in this table, including user, are dropped before the request reaches the model.

Response

The standard completion object. model is the model that actually answered. An extra drip field holds the request ID, the model you asked for, any fallback, and what the request cost.

{
  "id": "chatcmpl-…",
  "object": "chat.completion",
  "created": 1761040000,
  "model": "anthropic/claude-haiku-5.5",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Hello there, nice to meet you!" },
      "finish_reason": "stop"
    }
  ],
  "usage": { "prompt_tokens": 14, "completion_tokens": 8, "total_tokens": 22 },
  "drip": {
    "request_id": "req_3f9c…",
    "model": "anthropic/claude-haiku-5.5",
    "requested_model": "anthropic/claude-haiku-5.5",
    "fallback_from": null,
    "cost_micros": 6,
    "cost_usd": 0.000006
  }
}

Response headers

raw = client.chat.completions.with_raw_response.create(
    model="anthropic/claude-haiku-5.5",
    messages=[{"role": "user", "content": "Say hello in five words."}],
)
print(raw.headers.get("x-drip-model"), raw.headers.get("x-ratelimit-remaining-requests"))

reply = raw.parse()
drip = getattr(reply, "drip", None)  # extra field: model, cost_usd, request_id
print(reply.choices[0].message.content, drip)

Fallbacks

Each model lists fallback models (fallbacks in /v1/models). Claude Haiku 5.5 has none today, so if Anthropic fails or times out, the request fails with 503 all_upstreams_failed and you are not charged for it. When a model has fallbacks, Drip tries the next one your key may use, you pay for the model that answered, and x-drip-fallback-from shows that a fallback was used.

If the provider rejects the request itself (a 4xx other than 408, 409, 425 or 429), Drip returns that error with code upstream_rejected and does not try a fallback.

stream = client.chat.completions.create(
    model="anthropic/claude-haiku-5.5",
    messages=[{"role": "user", "content": "Write a haiku about rain."}],
    stream=True,
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
# The last chunk carries usage and a "drip" field with the model that answered and the cost.

After the last content chunk, Drip always sends one more chunk with empty choices, the token usage and the drip field, then data: [DONE]. You do not need stream_options to get usage.

data: {"id":"chatcmpl-…","object":"chat.completion.chunk","model":"anthropic/claude-haiku-5.5","choices":[{"index":0,"delta":{"content":" rain."}}]}

data: {"id":"chatcmpl-…","object":"chat.completion.chunk","model":"anthropic/claude-haiku-5.5","choices":[],"usage":{"prompt_tokens":15,"completion_tokens":19,"total_tokens":34},"drip":{"request_id":"req_…","model":"anthropic/claude-haiku-5.5","requested_model":"anthropic/claude-haiku-5.5","fallback_from":null,"cost_micros":12,"cost_usd":0.000012}}

data: [DONE]

• If the model stops partway, the stream ends with an error event with code stream_interrupted. You pay only for the tokens received.
• If you close the connection early, Drip stops the model and charges for the tokens generated until then.
• Drip waits for the first chunk before it starts the stream, so a provider that fails before answering returns a normal error instead of a broken stream.

Tool calls

Pass tools and read tool_calls from the reply. Run the tool yourself, then send its result back in a message with role tool.

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Current weather for a city",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    },
}]

reply = client.chat.completions.create(
    model="anthropic/claude-haiku-5.5",
    messages=[{"role": "user", "content": "Is it raining in Lisbon?"}],
    tools=tools,
)
call = reply.choices[0].message.tool_calls[0]
print(call.function.name, call.function.arguments)

JSON mode

Set response_format to {"type": "json_object"} and ask for JSON in the prompt.

import json

reply = client.chat.completions.create(
    model="anthropic/claude-haiku-5.5",
    response_format={"type": "json_object"},
    messages=[
        {"role": "system", "content": "Reply with JSON only."},
        {"role": "user", "content": "List three fruits with their colours."},
    ],
)
data = json.loads(reply.choices[0].message.content)

Images and files

Send images as image_url content parts, as an https URL or a base64 data URL. Send PDFs and text files as file parts with a base64 file_data URL. For the up-front credit check, each image or file part counts as about 1,500 tokens. You are billed for the tokens the model reports.

reply = client.chat.completions.create(
    model="anthropic/claude-haiku-5.5",  # needs "vision" in its capabilities
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "What is in this picture?"},
            {"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}},
        ],
    }],
)

Where credits come from

What a request costs

Cost = prompt tokens × input price + completion tokens × output price, at the prices of the model that answered. Prices include the service margin (currently 5%; the proposed range is 3–8%). Costs are counted in millionths of a dollar and rounded up.

How charging works

1. 01

   Hold

   Before calling the model, Drip holds the most the request could cost: the prompt plus max_tokens of output, at the model's prices.

2. 02

   Fit to your balance

   If your balance is lower than that, Drip lowers max_tokens to what it covers. If it covers fewer than 16 output tokens, the request fails with 402 insufficient_credits.

3. 03

   Settle

   After the reply, Drip charges the real cost from the token usage and releases the rest of the hold.

• See each request’s cost in drip.cost_usd in the response, and totals by key, model and day in the console.
• Unused credits never expire. They cannot be withdrawn as cash.
• All keys of an account share one balance.

Completion responses carry x-ratelimit-limit-requests, x-ratelimit-remaining-requests and x-ratelimit-reset-requests, so you can slow down before you reach the limit.

HTTP/1.1 429 Too Many Requests
retry-after: 12
x-ratelimit-limit-requests: 60
x-ratelimit-remaining-requests: 0
x-ratelimit-reset-requests: 12s

{"error":{"message":"Rate limit reached for API key 'research agent' (ending AbCd): 60 requests per minute. Try again in 12 s.","type":"rate_limit_error","param":null,"code":"rate_limit_exceeded"}}