API reference

Echo's API supports the OpenAI format and the Anthropic format. This page covers the endpoints (the URLs you call for each feature), authentication, streaming and the main parameters. For the details of requests and responses, follow each provider's API documentation.

Base URL and authentication

• The x-api-key header works on the Anthropic-format endpoints (/v1/messages, and /v1/models when it returns the Anthropic-format list). Other endpoints use the Authorization header
• The Anthropic format also requires the anthropic-version header (for example 2023-06-01). The Anthropic SDK sends it automatically
• Only HTTPS is accepted
• If the path you call has no /v1, you get HTTP 404 (the website's HTML page, not JSON) even with a correct key. If it has /v1 twice, you get HTTP 404 (a JSON Invalid URL error)

Base URL for each SDK and tool

Some SDKs need /v1 in the base URL and some add /v1 themselves. In the "Checked" column, "Yes" means we confirmed calls in Echo's test environment on 2026-09-27, and "No" means the value is based on the tool's official documentation.

Endpoints

Not available

Not tested

Model IDs

Put the ID of the model to call in model in the request. Echo model IDs are written the same way as the names used in the provider's API (for example deepseek-flash, claude-sonnet-5). The exact IDs are listed on each model's page in the model list.

You can also get the IDs of the models your key can use with GET /v1/models. With both the x-api-key and anthropic-version headers it returns the Anthropic-format list; otherwise it returns the OpenAI-format list.

curl "https://token.echohub.io/v1/models" \
  -H "Authorization: Bearer $ECHO_API_KEY"

Formats and models

You can call a model in a format different from its provider's. In that case, Echo's server converts the request and the response. After conversion, some parameters have no effect and some response fields differ from the provider's specification. We recommend calling each model in its provider's format.

What changes with conversion (checked in Echo's test environment on 2026-09-27):

This conversion also happens when you call a Claude model from a tool that only uses the OpenAI format, such as Cline, Cursor and Dify.

Call the Anthropic format with curl

curl "https://token.echohub.io/v1/messages" \
  -H "x-api-key: $ECHO_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "日本の首都はどこですか？"}]
  }'

Call the Responses API

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://token.echohub.io/v1",
    api_key=os.environ["ECHO_API_KEY"],
)

response = client.responses.create(
    model="gpt-6-sol",
    input="日本の首都はどこですか？",
)
print(response.output_text)

Streaming

Streaming means receiving the answer piece by piece while the model writes it. Add "stream": true to the request and the answer is returned as SSE (a mechanism for receiving data from the server one event at a time). Echo forwards the data from the upstream API in order, without buffering it.

What arrives when the upstream API returns an error or the connection is cut in the middle of an answer, and how to check, is in "When streaming stops partway" in Errors and fixes.

curl -N "https://token.echohub.io/v1/chat/completions" \
  -H "Authorization: Bearer $ECHO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-flash",
    "stream": true,
    "stream_options": {"include_usage": true},
    "messages": [{"role": "user", "content": "自己紹介を 3 文でしてください。"}]
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://token.echohub.io/v1",
    api_key=os.environ["ECHO_API_KEY"],
)

stream = client.chat.completions.create(
    model="deepseek-flash",
    messages=[{"role": "user", "content": "自己紹介を 3 文でしてください。"}],
    stream=True,
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://token.echohub.io/v1",
  apiKey: process.env.ECHO_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "deepseek-flash",
  messages: [{ role: "user", content: "自己紹介を 3 文でしてください。" }],
  stream: true,
});
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

import os
import anthropic

client = anthropic.Anthropic(
    base_url="https://token.echohub.io",
    api_key=os.environ["ECHO_API_KEY"],
)

with client.messages.stream(
    model="claude-sonnet-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "自己紹介を 3 文でしてください。"}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "https://token.echohub.io",
  apiKey: process.env.ECHO_API_KEY,
});

const stream = client.messages.stream({
  model: "claude-sonnet-5",
  max_tokens: 1024,
  messages: [{ role: "user", content: "自己紹介を 3 文でしてください。" }],
});
stream.on("text", (text) => {
  process.stdout.write(text);
});
await stream.finalMessage();

Main parameters

The available parameters and value ranges differ by model. Follow each provider's documentation.

Chat Completions (POST /v1/chat/completions)

Messages (POST /v1/messages)

Echo removes three parameters, inference_geo, service_tier and speed, before passing the request to the upstream. The processing region is fixed for each model (Data handling).

Responses (POST /v1/responses)

Embeddings

curl "https://token.echohub.io/v1/embeddings" \
  -H "Authorization: Bearer $ECHO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "[埋め込みモデルの ID]", "input": "東京は日本の首都です。"}'

You can find models that support embeddings in the model list. Both number arrays and base64 are supported.

Balance, usage and limits

• Check your balance and usage in the console (API keys and usage). There is currently no way to get your balance through the API
• Per-key spending limits, rate limits and the request size limit are in Limits and rate limits

Sources