One key, every model, never a token charge.

A gateway that runs on your own provider keys, a chat that does real research, and a computer that writes code and ships it live. Your provider bills the tokens. Credits pay only for the computer, the search, the reading and the indexing.

Three services on one account, with zero token markup

Everything wired into one gateway.

One address in front of every provider you use. Your keys stay yours, your model names stay stable, and each request goes home to whoever bills it.

One endpoint

OpenAI and Anthropic requests answer at the same address. Point any SDK at it and keep the rest of your code.

api.nemu.cclive
POST/v1/chat/completionsanthropic200 · 1.2s
POST/v1/messagesopenai200 · 0.8s
POST/v1/embeddingsgoogle200 · 0.3s
GET/v1/models42 mapped200 · 45ms
POST/reader?url=…markdown200 · 2.1s
POST/v1/chat/completionsquota rule429 · paused

Your keys

Encrypted at rest and billed by the provider that issued them.

  • openai-prodsk-••••4f2a
  • anthropic-mainsk-••••c91d
  • mistral-eusk-••••7b30

Mappings

The names your code asks for, and the provider model each one resolves to right now.

  • smartclaude via anthropic-main
  • fastgpt-mini via openai-prod
  • cheapmistral via mistral-eu
14 ms median gateway overhead42 models mapped across five providersSee what the console shows

Stop configuring. Start shipping.

One base URL, zero rewrites

Your existing OpenAI or Anthropic code keeps working. Swap the address, keep everything else.

Both wire formats, interchangeable

Call any mapped model from either SDK. The gateway translates between formats on the fly.

Your aliases, not vendor names

Map "fast", "smart" or "cheap" to any provider model and swap what's behind them without a deploy.

app.py
from openai import OpenAI
client = OpenAI(
base_url="https://api.nemu.cc/v1",
api_key=NEMU_KEY,
)
reply = client.chat.completions.create(
model="smart",
messages=[{"role": "user", "content": "hi"}],
)

Aliases like “smart” and “fast” are mappings you define once in the console, and every platform picks them up instantly.

Everything that speaks OpenAI already speaks nemu.

You do not have to write the integration. Any tool that lets you set a base URL points at the gateway and inherits the whole account: your keys, your model names, your quotas, your usage record.

OPENAI_BASE_URLhttps://api.nemu.cc/v1
ANTHROPIC_BASE_URLhttps://api.nemu.cc
  • Cursoreditor
  • Clinecoding agent
  • Continueeditor
  • Zededitor
  • Aiderterminal
  • Vercel AI SDKframework
  • LangChainframework
  • LlamaIndexframework
  • Open WebUIchat front end
  • Raycastdesktop
  • n8nautomation
  • Your own servicewhatever you wrote

None of these need a plugin, an adapter or a proxy of their own. They take an address, and this is the address.

Drag a day across. The tokens never reach us.

Tokens are between you and whoever issued the key. What we charge for is throughput, and the work nemu goes and does: searching, reading, running the computer, building an index. Push the traffic up and watch which of the two actually moves.

28M tokens a day
1M a day1B a day
Tokens a day, over a monthmillions, on your own keys
Managed by you

Every token above is billed by the provider whose key it ran on, at your rates, on your existing account. It never touches an invoice of ours, and we cannot see what you pay.

your contract
Metered by nemuFree€0

The free account carries this on its own: 15 requests a minute, with fixed daily allowances instead of a credit balance.

What nemu does for you, fixed on Freeno credits involved
  • Web searches25 a day
  • Page reads25 a day
  • Deep research3 runs a day
  • Document uploads5 a day
Fixed allowances, no credit balance and no card€0.00
Tokens charged by nemu, at 28M tokens a day€0.00

The wait should be your provider, not us.

Your request lands on whichever edge is closest, gets resolved there, and leaves for your provider immediately. The only slow part of a call should be the model thinking, so we keep our own share of it small enough to stop noticing. Hover a country to see what sits in it.

Where the edge sits today

6 live · 2 on the way
Live now

Los Angeles, Ashburn, London Docklands, Frankfurt, Wyszków, Bucharest.

Next

Paris, then a first region in Asia. Neither is live yet, so nothing routes there.

€0

Everything unlocked.

Free is one person, with daily allowances instead of a bill and your own models on your own keys from the first minute. Anything paid from credits, the computer included, starts on Starter.

Create your account
Day one on Freeno payment step
  1. 1
    Connect a provider key
    OpenAI, Anthropic and more, encrypted at rest
  2. 2
    Map your first model alias
    call it "fast" and point any SDK at it
  3. 3
    Search and read the web
    25 searches and 25 page reads, every day
  4. 4
    Run deep research
    3 full runs a day, citations included
  5. 5
    Open a second workspace
    2 of your own, from day one

Plans that meter
services, not tokens.

Compare all features →

Free

A solo testing sandbox for the gateway, the chat and the reader.

€0

/month

Get started

Starter

Individuals and small projects moving past evaluation.

€5.99

/month, incl. €8 platform credits

Get started

Pro

Active teams and multi-app workloads.

€19.99

/month, incl. €25 platform credits

Get started

Max

Heavy teams and small companies running at scale.

€59.99

/month, incl. €90 platform credits

Get started

Enterprise

Large organizations with many apps and users.

Custom

Billed annually

Contact sales
Get started now

Start with your keys, keep the whole bill.

The gateway, the chat and the computer on one account. Free is enough to evaluate the gateway and the chat on your own keys, and your provider keeps billing the tokens exactly as it does today.

No token markup. Credits only pay for what nemu runs