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

# Platform architecture

> How applications, Neurons, trust services, and the Neuro-Ledger fit together

Applications connect to a **Neuron**, the operator-hosted runtime that exposes Neuro capabilities. Choose an integration based on where your code runs and who manages credentials and private keys.

```text theme={null}
Application backend ── SDK or HTTPS Agent API ──┐
AI application ─────── MCP tools ───────────────┤
Direct client ──────── XMPP ────────────────────┤
                                                 ↓
                                              Neuron
                                                 │
                  Accounts, identities, contracts, features, wallets
                                                 │
                                                 ↓
                                           Neuro-Ledger
```

The diagram is a developer navigation model. It does not imply that every Neuron enables every capability or that every deployment shares one central API host.

## Choose where credentials are managed

| Caller                 | Recommended interface                                 | What the caller manages                                         |
| ---------------------- | ----------------------------------------------------- | --------------------------------------------------------------- |
| Backend service        | Supported SDK, otherwise Agent API                    | Application credentials, sessions, and secure server-side state |
| Browser or mobile UI   | Application backend plus a supported user-facing flow | User interaction; no embedded Agent API secret                  |
| AI application         | Supported, scoped MCP tools                           | Tool selection and human approval rules                         |
| Direct protocol client | XMPP                                                  | Its connection and private keys                                 |
| Neuron extension       | Neuron package                                        | Package lifecycle and operator-approved privileges              |

The **Agent API** is the Neuron's main HTTP API. “Agent” means the application acts through the Agent endpoint rather than connecting directly over XMPP while holding the corresponding private keys. It is unrelated to AI agents.

## Platform objects

A typical trusted workflow crosses multiple layers:

```text theme={null}
Account → key → Legal Identity → contract role → signature
                                      │
                                      ├→ Neuro-Feature → state → history
                                      └→ wallet operation → transaction state
```

Read the [object model](/platform/object-model) before designing persistence or lifecycle handling.

## Federation and deployment

A Neuron belongs to an operator-controlled environment. Cross-domain behavior depends on the deployed services, privileges, trust relationships, and federation configuration. Applications must receive an exact Neuron host and an environment contract from the operator; they must not infer production reach from a successful local request.

## Choose what to read next

* [Choose an integration](/get-started/choose-integration)
* [Prepare a development environment](/get-started/environments)
* [Trust and federation](/platform/trust-and-federation)