alltoken

Official Python SDK for AllToken — one API for OpenAI, Anthropic, and 100+ models.

pip install alltoken

Requires Python 3.10+.

Quick start

from alltoken import AllToken

client = AllToken(api_key="...")  # or os.environ["ALLTOKEN_API_KEY"]

# OpenAI-compatible surface (maps to /v1)
resp = client.openai.raw.post(
    "/chat/completions",
    json={
        "model": "gpt-4o",
        "messages": [{"role": "user", "content": "Hello!"}],
    },
)
print(resp.json())

# Anthropic-compatible surface (maps to /anthropic)
resp = client.anthropic.raw.post(
    "/messages",
    json={
        "model": "claude-sonnet-4",
        "max_tokens": 1024,
        "messages": [{"role": "user", "content": "Hello!"}],
    },
)
print(resp.json())

The same API key works for both surfaces. Model catalog: alltoken.ai/models.

Configuration

AllToken(
    api_key="...",                      # required
    base_url="https://api.alltoken.ai", # optional, defaults to production
    default_headers={"X-My-Tag": "a"},  # optional, merged into every request
)

API surface

Field	Spec	Base URL
`client.openai.raw`	`chat.yml` (OpenAI-compatible)	`https://api.alltoken.ai/v1`
`client.anthropic.raw`	`anthropic.yml`	`https://api.alltoken.ai/anthropic`

.raw is a pre-configured httpx.Client — base URL + auth are set, call .get() / .post() / .stream() directly. Pydantic models for request/response bodies are generated from the OpenAPI specs into alltoken.generated.chat and alltoken.generated.anthropic.

Status

v0.1.0 — Scaffold. Pydantic models are generated from the spec, the wrapper surface is minimal. Expect breaking changes in 0.x. Ergonomic helpers (client.chat.completions.create(...), async streaming iterators, retries, etc.) are coming in 0.2.x.

Contributing / Local development

# Clone megaopenrouter as a sibling (for the OpenAPI specs)
git clone git@gitlab.53site.com:ai-innovation-lab/megaopenrouter.git ../megaopenrouter

# Install with dev deps
pip install -e ".[dev]"

# Regenerate pydantic models from specs
python scripts/generate.py

# Test + lint
pytest
ruff check
mypy .

Generated models live in src/alltoken/generated/{chat,anthropic}.py — these are committed so users who install from PyPI don't need to run codegen.

License

MIT

Metadata

Release files for alltoken 0.2.33

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for alltoken 0.2.33
File	Size	Uploaded
alltoken-0.2.33.tar.gz	25.7 kB	Aug 13, 2026	Details

Built distribution (wheel)

Table of built distributions (wheels) for alltoken 0.2.33
File	Interpreter	ABI	Platform	Reset
alltoken-0.2.33-py3-none-any.whl 29.0 kB Aug 13, 2026	Python 3	none	any	Details

Total release size: 54.6 kB

Release files / alltoken-0.2.33.tar.gz

Download URL	alltoken-0.2.33.tar.gz
Size	25.7 kB
Tags	Source
SHA-256 checksum How to use checksums	`1e937d3a65c51b9f64fa03dc7ef6cde50d91651980569c3f14a3c2ea404ee957`
BLAKE2b-256 checksum How to use checksums	`f5c2bc465183e1c97d196eb729107f38cccb7b95d41d0db4d90da0b47613260d`
Upload date	Aug 13, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.12.13`

Release files / alltoken-0.2.33-py3-none-any.whl

Download URL	alltoken-0.2.33-py3-none-any.whl
Size	29.0 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`8d39f388692eebb73720a4b744554295802e27bb83c81020c5ecd4f58e1440d2`
BLAKE2b-256 checksum How to use checksums	`eb9398223eda0e706d9c720b0aa60c3750beb041b1b4827357ad7bbfe53413a8`
Upload date	Aug 13, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/7.0.0 CPython/3.12.13`