← Voltar aos guias

Call Claude Messages API via 59API: cURL and Python

Build a Claude Messages API Request with 59API

The Claude Messages API is Anthropic’s conversational endpoint for sending a list of messages and receiving structured model output. If you already use Claude-compatible tooling, 59API lets you route those requests through https://api.59api.com while keeping the familiar Messages API format. This is useful when you want pay-as-you-go access to Claude Opus, Sonnet, Haiku, and Fable models without changing your application architecture.

In this hands-on example, you will send one prompt with cURL, then make the same request using the Anthropic Python SDK. Before starting, create a 59API account, generate an API key, and store it in an environment variable rather than hard-coding it into a repository. On macOS or Linux, run: export API_KEY="your_59api_key".

Know the Messages API Request Shape

A Messages API call is a POST request to https://api.59api.com/v1/messages. It needs an API key, the Anthropic API version header, a model identifier, a token limit, and a messages array. Each message has a role such as user or assistant and content. Use the exact model ID shown in your 59API dashboard or model catalog; model availability and versioned IDs can change over time.

• model: The Claude model to run, such as an enabled Sonnet model ID.
• max_tokens: The maximum number of completion tokens to generate.
• messages: The ordered chat history sent to the model.
• system: An optional top-level instruction for behavior, formatting, or tone.

Send Your First Request with cURL

This request asks Claude for a concise explanation of rate limiting. Replace the example model ID if your 59API catalog exposes a different current identifier.

curl https://api.59api.com/v1/messages -X POST -H "x-api-key: $API_KEY" -H "anthropic-version: 2023-06-01" -H "content-type: application/json" --data '{"model":"claude-sonnet-4-5-20250929","max_tokens":300,"system":"Answer clearly for a backend developer.","messages":[{"role":"user","content":"Explain API rate limiting in three bullet points."}]}'

A successful response includes an id, a stop_reason, token usage data, and a content array. For a simple text answer, the generated text is usually in the first content block. Do not assume every response is plain text, however: tool-use and other block types can appear in more advanced integrations. Production code should inspect each content item’s type before displaying or processing it.

Use the Anthropic Python SDK Without Rewriting Your App

Because 59API supports the Claude-compatible API pattern, you can point the official-style Python client at its base URL. First install the package with pip install anthropic, then use the following code:

from anthropic import Anthropic

client = Anthropic(api_key="your_59api_key", base_url="https://api.59api.com")

message = client.messages.create(
    model="claude-sonnet-4-5-20250929",
    max_tokens=300,
    system="Answer clearly for a backend developer.",
    messages=[{"role": "user", "content": "Give me one Python retry strategy for HTTP 429 errors."}]
)

print(message.content[0].text)

Keep the API key in an environment variable in real deployments. For example, load it with os.environ["API_KEY"]. This avoids leaking credentials through source control, screenshots, or browser bundles. Use server-side code or a secure backend proxy for web applications; never expose a long-lived AI API key in frontend JavaScript.

Choose Models and Control Spend Deliberately

Model selection has a larger cost and latency impact than small prompt edits. Use a fast, economical Haiku-class model for classification, extraction, routing, and short summaries. Choose Sonnet for most coding and reasoning tasks, and reserve Opus-class models for difficult analysis where the additional capability is justified. Set a realistic max_tokens value, because overly generous limits can increase spend and response time.

59API is a strong low-cost option for teams that need native, official-quality Claude model access without a downgraded substitute. Its pay-as-you-go relay model also makes it practical to test several Claude families before committing a workload to one. If you share the service with other developers, check its referral rebate program as an additional way to reduce effective API costs.

Handle Streaming, Limits, and Failures

For chat interfaces, add "stream": true to the request body and consume server-sent events as they arrive. Streaming improves perceived speed because users can read output before the full response is complete. For background jobs, non-streaming requests are often simpler because one completed JSON response can be stored and processed atomically.

Finally, handle HTTP 429 and transient 5xx responses with capped exponential backoff plus random jitter. Respect any retry guidance returned by the service, log request IDs when available, and avoid blindly retrying user-visible generation requests that could create duplicate outputs or charges. Start by signing up for 59API, make the cURL request above, and then move the same Messages API call into your preferred SDK.

Preços · Docs