> ## Documentation Index > Fetch the complete documentation index at: https://docs.autogrowth.farm/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > AutoGrowth runs social accounts on real phones. Everything a page describes happens on a physical device in the customer's own fleet, never through a social platform's API. > The Companion is a macOS desktop app that runs on the customer's own Mac and manages the phones over USB — iPhones through WebDriverAgent, GrapheneOS phones through ADB and scrcpy. Nothing runs while the Companion is closed. > The Companion builds, signs and installs WebDriverAgent on each iPhone itself. Never tell a customer to sign WebDriverAgent by hand in Xcode; Xcode is a prerequisite the Companion drives, not a step the customer performs. > Autogrowth MAX is a one-time, invite-only lifetime license bought from the organization's credit balance, not a subscription tier. It is required to run any automation, workflow or Incubator program. Without a price set for that organization, the purchase is refused. > An automation is one of AutoGrowth's built-in definitions; a script is a graph the customer builds in the Script Builder. The Script Builder is locked platform-wide, so never present building a script as something a customer can do today. > An API key is a bare 64-character hex string with no prefix. It is shown once, at creation, is sent as `Authorization: Bearer `, and can never be read back. > An API key sees only the key owner's own accounts, content, collections, prompts, scripts and executions. Devices and tags are organization-scoped, and a non-owner sees only the devices assigned to them or carrying one of their tags. The automation catalog is global. > There is no pagination. List endpoints return the whole set; the only limiting parameter anywhere is `limit` on `GET /v1/executions`, which defaults to 50 and has no offset to page with. > Remote control and the live screen are dashboard-only, over WebSocket and MJPEG. Neither is part of the public API. > The MCP server is hosted at https://mcp.autogrowth.farm and authenticates with OAuth. It is not an npm package and there is nothing to install locally. Its tools inherit exactly the scope of the API key behind them. > Android support is partial and GrapheneOS-only: those phones appear in the fleet and can be watched and remote-controlled, but they cannot run scripts or automations. > Never document Hi Katie or any persona feature. When a page and the code disagree, the code wins — check a claim against the source before repeating it. # MCP tools > The tools the server exposes for listing devices, accounts, automations and media, and for scheduling runs. Eleven tools. Five read, four schedule, two handle media. | Tool | What it does | Key inputs | | --------------------------- | ----------------------------------------------------------- | --------------------------------------------- | | `list_automations` | The automation catalog, with each automation's input schema | none | | `schedule_automation` | Queue one automation on one phone | `automationId`, `deviceId`, `scheduledAt` | | `bulk_schedule_automations` | Queue up to 100 automation runs in one call | `executions[]` | | `list_scripts` | Your scripts | none | | `schedule_script` | Queue one script on one phone | `scriptId`, `deviceId`, `scheduledAt` | | `bulk_schedule_scripts` | Queue up to 100 script runs in one call | `executions[]` | | `list_schedules` | Scheduled and finished runs | `deviceId`, `scriptId`, `from`, `to`, `limit` | | `list_devices` | Your phones and their live state | none | | `list_accounts` | Your social accounts | `platform` | | `upload_media` | Add a file to the content library from a public URL | `url` | | `list_media` | Browse library metadata | `folder`, `starred` | Every tool runs against the [REST API](/api/overview) as you. If a tool returns less than you expect, check [what a key can see](/api/authentication#what-a-key-can-see) before assuming a bug. ## Reading ### list\_automations No parameters. Returns every automation in the catalog: `id`, `name`, `description`, `category`, `estimatedDuration` in minutes, and an `inputs` array. `inputs` is the schema for `inputValues` when you schedule. Each entry has an `id` (the key to use), a `type`, a display `name`, sometimes `options` and a `defaultValue`. **Call this before scheduling anything** — automation IDs and their inputs are not guessable, and an assistant that invents them gets a 404. ### list\_scripts No parameters. Returns your scripts with their IDs, names, descriptions and estimated durations. Scripts with an empty graph are omitted. The Script Builder is locked platform-wide, so most fleets have no scripts and this returns an empty list. Automations are the thing to schedule. ### list\_devices No parameters. Returns every phone you can reach — an owner sees the whole organization, anyone else sees phones assigned to them or carrying one of their tags — with `id`, `name`, model, OS, `online`, `status`, `battery`, `slotState`, tags, and the accounts on the phone. The `id` is what `deviceId` wants. `slotState` must be `active` for a run to be accepted. ### list\_accounts | Parameter | Type | Notes | | ---------- | ----------------------- | -------------------------------- | | `platform` | `instagram` \| `tiktok` | Optional. Omit for all accounts. | Returns the accounts **you** own, with username, platform, status, follower counts and growth. The `username` is the value an `account_selection` input wants — pass it back exactly as returned. Usernames are stored lowercase and without a leading `@`, whatever an assistant's example shows. The filter only accepts `instagram` and `tiktok`. X and Threads accounts exist in AutoGrowth and are returned when you omit `platform`, but they cannot be filtered for. ### list\_schedules | Parameter | Type | Notes | | ---------- | ------ | ------------------------------------------ | | `deviceId` | string | Only runs on this phone | | `scriptId` | string | Only runs of this script | | `from` | string | ISO 8601 lower bound on the scheduled time | | `to` | string | ISO 8601 upper bound | | `limit` | number | Defaults to 50 | Returns executions newest scheduled first. The device, script, account and container references are expanded into objects in place; `automationId` stays a plain string. Each carries a `status` — `pending`, `processing`, `completed`, `failed`, `skipped` or `cancelled`. There is no offset, so `limit` is a cap and not a page. Raising it on a busy fleet returns a lot of JSON; see [limitations](/mcp/limitations). ## Scheduling ### schedule\_automation | Parameter | Type | Notes | | ------------------ | ------- | ---------------------------------------------------------------------------------------- | | `automationId` | string | Required. From `list_automations`. | | `deviceId` | string | Required. From `list_devices`. | | `scheduledAt` | string | Required. ISO 8601, e.g. `2026-09-01T14:00:00Z`. | | `inputValues` | object | Keyed by input `id`. See the table below. | | `enableAIRecovery` | boolean | Let the Companion recover from an unexpected screen. | | `switchMobileData` | boolean | Cycle cellular data off and on first, for a fresh carrier IP. Only `true` does anything. | ### schedule\_script Identical, with `scriptId` in place of `automationId`. A script's `inputValues` keys are node IDs from its graph, not friendly names. ### Building inputValues Use each input's `id` as the key. The value's shape depends on the input's `type`: | Input type | Value | | ------------------- | ---------------------------------------------------------------------------------------------------------------------- | | `account_selection` | The account's username, as a string | | `user_input` | A string | | `user_toggle` | A boolean | | `user_select` | One of that input's `options`, as a string | | `user_multi_select` | An array of `options` values | | `push_media` | An array of content IDs from `list_media` | | `container_select` | An existing container's **name**, as a string | | `container_target` | A container ID, optionally with sibling keys `containerName` and `proxyId` when a fresh container is being provisioned | The tool's own description only teaches an assistant the first six types. If an automation asks for `container_select` or `container_target`, say so explicitly in your prompt, or the assistant will guess. ### bulk\_schedule\_automations and bulk\_schedule\_scripts Both take a single `executions` array of 1 to 100 items. Each item carries its own `deviceId`, `scheduledAt`, `inputValues`, `enableAIRecovery` and `switchMobileData`, plus `automationId` or `scriptId` — so one call can put a different account and different media on every phone. Items are validated independently. The response reports `created` and `failed`, with each failure naming its `deviceId`, an error and sometimes a code such as `MAX_REQUIRED` or `SLOT_EXPIRED`. **A call where every item failed is still a success at the protocol level**, so read the counts rather than trusting "done". ### What scheduling rejects | Cause | What comes back | | ------------------------------------------------ | ------------------------------- | | No Autogrowth MAX, and the run is an automation | `MAX_REQUIRED` | | The phone's slot has expired | `SLOT_EXPIRED` | | The create-account automation without its add-on | `CREATE_ACCOUNT_ADDON_REQUIRED` | | Unknown automation, script, device or account | Not found | The codes are explained in [Errors](/api/errors). Overlapping runs are **not** skipped. A phone runs one execution at a time and queues the rest, so scheduling two things at 14:00 gives you two runs, the second starting late. Some assistants claim a one-minute conflict rule exists; it does not. ## Media ### upload\_media | Parameter | Type | Notes | | --------- | ----------------------------------------------------- | ----------------------------------------- | | `url` | string | Required. A **publicly reachable** URL. | | `title` | string | Display title. | | `type` | `image` \| `video` \| `carousel` \| `reel` \| `story` | Detected from the file when omitted. | | `folder` | string | Legacy folder name. Defaults to `images`. | The server fetches the URL itself, so a signed link, a localhost address or anything behind a login fails. The file name is taken from the URL. The response includes the new content ID, which is what `push_media` inputs want. An upload that would exceed your organization's storage allowance is rejected with `STORAGE_LIMIT_EXCEEDED`. There is no way to file the upload into a collection through this tool — see [limitations](/mcp/limitations). ### list\_media | Parameter | Type | Notes | | --------- | ------- | ----------------------------------------------------- | | `folder` | string | Filter by legacy folder. Pass `all` for every folder. | | `starred` | boolean | `true` returns only starred items. | Returns metadata only: ID, title, type, folder, size, dimensions, status, usage count and dates. **File and thumbnail URLs are stripped**, so an assistant can list your media but cannot look at it or link to it.