> ## Documentation Index
> Fetch the complete documentation index at: https://docs.e6data.com/llms.txt
> Use this file to discover all available pages before exploring further.

# HTTP API

> http_api

ReferenceApplies to: Unified engine deployments

`[http_api]`

## Settings

### `http_api`

Configures the HTTP service used by the browser UI and HTTP clients.

* **Default:** Fields below
* **Type:** object; optional

> **Example**
>
> A native process can serve the UI and API at `http://127.0.0.1:15003`. In the combined platform host, Unified uses the shared host HTTP listener instead.

### `owner`

Chooses the user identity that the HTTP service assigns to its requests.

* **Default:** `"local"`
* **Type:** string

> **Example**
>
> `"local"` makes HTTP-originated sessions belong to the local identity. This value is chosen by the listener rather than taken from a client’s claimed identity.

### `max_request_bytes`

Limits the encoded size of one incoming request.

* **Default:** `16777216`
* **Type:** integer (bytes); positive

> **Example**
>
> `16777216` accepts a request up to 16 MiB. A large submitted plan or SQL request counts toward that request’s limit.

### `max_response_bytes`

Limits the encoded size of one outgoing response.

* **Default:** `134217728`
* **Type:** integer (bytes); positive

> **Example**
>
> `67108864` gives one response a 64 MiB budget. A larger result can require smaller result pages rather than one larger response.

### `max_page_size`

Limits how many records one list response may return.

* **Default:** `100`
* **Type:** integer; positive

> **Example**
>
> `100` returns a long list over pages of at most 100 items. A client can still request a smaller page.

### `notebook_poll_interval`

Sets how often an HTTP notebook request checks execution progress while waiting.

* **Default:** `100 ms`
* **Type:** duration

> **Example**
>
> A 100 ms interval checks about ten times per second. This does not set how often a notebook is scheduled to run.

### `shutdown_timeout`

Sets how long the HTTP server has to finish existing requests during shutdown.

* **Default:** `5 s`
* **Type:** duration

> **Example**
>
> `{ secs = 5, nanos = 0 }` gives active HTTP requests a five-second drain period.

## MCP endpoint

### `mcp.enabled`

Chooses whether the HTTP service exposes its Model Context Protocol endpoint for tool clients.

* **Default:** `true`
* **Type:** boolean

> **Example**
>
> `false` turns off MCP access while leaving ordinary HTTP API routes available. It does not stop the whole HTTP server.

### `mcp.allowed_hosts`

Lists the hostnames that MCP requests are allowed to target.

* **Default:** `["localhost", "127.0.0.1", "::1"]`
* **Type:** array

> **Example**
>
> For a local service, entries such as `"localhost"` and `"127.0.0.1"` allow those Host header names. List hostnames here, not complete browser origin URLs.

### `mcp.allowed_origins`

Lists browser origins allowed by the MCP endpoint’s origin check.

* **Default:** `[]`
* **Type:** array

> **Example**
>
> An origin includes scheme, host, and optional port, such as `"https://console.example.com"`. An empty list does not add an exception for a browser origin.

### `mcp.max_response_bytes`

Limits the encoded size of one outgoing response.

* **Default:** `1048576`
* **Type:** integer (bytes); positive

> **Example**
>
> `67108864` gives one response a 64 MiB budget. A larger result can require smaller result pages rather than one larger response.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.