> ## Documentation Index
> Fetch the complete documentation index at: https://docs.nemu.cc/llms.txt
> Use this file to discover all available pages before exploring further.

# How nemu is put together

> Three surfaces over one account, and what each is for

nemu is three products sharing one account, one balance and one set of
credentials.

| Surface | Where        | What it is                                             |
| ------- | ------------ | ------------------------------------------------------ |
| Console | app.nemu.cc  | Where you configure providers, models, keys and limits |
| Gateway | api.nemu.cc  | The API your code calls                                |
| Chat    | chat.nemu.cc | The assistant, and the Computer it works in            |

You sign in once at id.nemu.cc and all three accept that session.

## The idea

You bring your own provider keys. nemu holds them, decides which model a request
should reach, enforces your limits, records what happened, and forwards it. Your
applications only ever see a nemu key, so a provider key never leaves the
console and rotating one is a single edit rather than a redeploy.

That is the gateway. Chat and the Computer are built on top of the same
machinery, so a conversation and a production request are metered, limited and
logged the same way.

## The console

app.nemu.cc is where the configuration lives.

You add a **provider**, which is somewhere models come from, either picked from
the models.dev catalog or defined by hand with your own base URL. Under it you
enable **models**, each with a gateway name of your choosing. You create
**API keys** for your applications. Everything else is optional and shapes how
requests behave: **mappings** rewrite one model name to another, **prompts**
inject a system prompt for a given model, **quotas** cap usage per rule,
**workflows** route a request through logic before it reaches a model, and
**indexes** make a website searchable.

**Metrics** and **Usage** show what actually happened, per request and in
aggregate.

## The gateway

api.nemu.cc speaks three protocols: OpenAI chat completions, the OpenAI
Responses API, and Anthropic Messages. A request in any of them can be served by
a provider that speaks a different one, because nemu translates in both
directions.

Read [How the gateway works](/setup/gateway) for the endpoints and the rules a
client has to follow, and the setup guides for
[Claude Code](/setup/claude-code) and [Codex](/setup/codex).

## Chat

chat.nemu.cc is an assistant that uses the models you configured. It searches
the web, reads your files, connects to your databases and servers, and can hand
work to subagents that run in parallel. It also has the
[Computer](/computer), a real machine it can write code in and put the result
online.

Chat runs its work as jobs rather than inside the browser, so closing the tab
does not stop it.

## Workspaces

Everything above belongs to a **workspace**. Providers, models, keys, mappings,
quotas and workflows are all scoped to one, so a key issued in one workspace
cannot reach a model enabled in another.

Use a second workspace to keep environments apart, or to give a team its own
set of providers without touching yours.

## What it costs

Two separate things.

Your **plan** sets the limits: how many requests, how many providers, how much
of the platform is switched on. See [Plans and pricing](/plans).

Your **credits** pay for work nemu does for you: minutes on the Computer, web
searches, page reads, indexing. Model tokens are never part of this. Those are
billed by whichever provider you connected, on your own key.