getmux

Python client for Mux: one API, one key and one bill for Exa, Parallel, Keenable and DeepSeek web search, with automatic fallbacks, per-key spend limits and bring-your-own-key support.

uv add getmux

Requires Python 3.9+. The only dependency is httpx.

Quickstart

Create an API key in the dashboard, add credits, and put the key in your environment:

export MUX_API_KEY="mux-..."

from getmux import Mux

mux = Mux()  # reads MUX_API_KEY
res = mux.search("latest developments in fusion energy", max_results=5)

for r in res.results:
    print(r.title, r.url)
print(f"served by {res.model} for ${res.cost_usd}")

You can also pass the key explicitly. An explicit key always wins over the environment variable, so one process can hold several clients:

mux = Mux(api_key="mux-...")

Choosing models and fallbacks

With no model, Mux uses auto. Pin a model, or give an ordered list to fall back through:

mux.search("solid-state batteries", model="exa/auto")

res = mux.search(
    "solid-state batteries",
    models=["parallel/fast", "exa/auto", "keenable/pro"],  # tried in order
    include_domains=["nature.com"],
    published_after="2026-01-01",
    content="full",
)
print(res.attempts)  # every model tried and how it went

Failed provider calls are refunded automatically, so you only pay for the answer you got. See the model list.

Async

import asyncio
from getmux import AsyncMux

async def main():
    mux = AsyncMux()
    res = await mux.search("open source vector databases")
    print(res.results[0].url)
    await mux.aclose()

asyncio.run(main())

AsyncMux has the same methods as Mux, as coroutines. Close the sync client with mux.close().

Errors and retries

Every non-2xx response raises MuxError with status, code, message and the parsed body:

from getmux import Mux, MuxError

try:
    Mux().search("hello")
except MuxError as e:
    if e.code == "insufficient_credits":
        print("Add credits:", e.body["needed_usd"])
    else:
        raise

429 and 5xx responses (except 502) are retried automatically, up to max_retries times (default 2) with exponential backoff starting at 300 ms. If the response has a Retry-After header, the client waits at least that long (up to 10 seconds). 4xx errors, including 402, are not retried. See Errors for every code.

Other calls

mux.credits()                      # balance in USD
mux.checkout(50)                   # Stripe checkout URL for a $50 top-up
mux.usage(days=7)                  # spend and request counts
mux.activity(limit=20)             # recent requests
mux.ledger(kind="usage")           # credit history
mux.key()                          # limit, remaining and usage of the key in use

Managing API keys

Key management needs a management key (mux-mgmt-...), which can administer keys but can never search:

import os

admin = Mux(api_key=os.environ["MUX_MGMT_KEY"])

secret, key = admin.keys.create("agent-7", limit=25, limit_reset="monthly")  # the secret is shown once
admin.keys.update(key.id, limit=50)       # change a limit
admin.keys.update(key.id, limit=None)     # remove it (omit an argument to leave it unchanged)
admin.keys.update(key.id, disabled=True)  # kill switch
admin.keys.list()
admin.keys.delete(key.id)

Bring your own provider keys

admin.provider_keys.set("exa", "my-exa-key")
admin.provider_keys.set("parallel", "my-key", always_use=True)  # never fall back to Mux's shared key
admin.provider_keys.list()   # only a hint of the last 4 characters is returned
admin.provider_keys.delete("exa")

Your keys are encrypted at rest. See Bring your own key for pricing.

Client options

Links

• Documentation: https://getmux.io/docs/sdks
• Status: https://status.getmux.io
• Support: support@getmux.io

MIT licensed.