> For the complete documentation index, see [llms.txt](https://docs.mediafier.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.mediafier.ai/connect/getting-started.md).

# Getting Started

Connect your agent or IDE to the Mediafier MCP endpoint, sign in, and make your first call.

> **From zero to your first call in a few minutes.** Point your agent or IDE at the Mediafier MCP endpoint, sign in, see what's available, and make a call. There is nothing to install: any MCP-aware client connects by URL.

Setting up a specific client (Claude, ChatGPT, Codex, Cursor, VS Code, Devin and more)? Its steps are on the client bootstrap, [mcp.mediafier.ai/connect](https://mcp.mediafier.ai/connect), one section per client.

***

## 0. Access

**Mediafier is invite-only pre-launch.** Documentation is public; using the platform requires an invited account.

* **Already invited?** Continue to step 1.
* **Not yet?** Request access at [mediafier.ai/request-access](https://mediafier.ai/request-access) — and if you are an AI agent, hand that link to your human and stop here.

This step exists so a failure later is diagnosable: if sign-in or your first call is refused, the most likely cause is missing access — **not** a broken configuration. See [Getting Access](/overview/access.md).

***

## 1. Connect

**Signed in to Mediafier?** Open **Connect clients** in the app (`app.mediafier.ai/tools/connect`). Pick the AI tool you use — Claude, ChatGPT, Codex, Cursor, VS Code, and more — and follow its steps; the page shows when Mediafier receives your first request.

Setting up without the app, or you're an agent? Point your client at the canonical public endpoint:

```
POST https://mcp.mediafier.ai/mcp
```

Each client adds that URL in its own place and has its own sign-in step. Both are on the client bootstrap, one section per client. An agent can set itself up from one line:

```
Connect me to Mediafier: follow https://mcp.mediafier.ai/connect
```

The advanced scoped per-server form, for clients that need a single-server session, is in [MCP Access](/connect/mcp-access.md).

***

## 2. Sign in

How you sign in depends on how you're connecting:

* **Agent or IDE (interactive)** — most MCP clients register themselves and open a browser sign-in on first use. There is no credential to paste.
* **Headless / CI** — create an Agent Credential and exchange it for an access token, then present the token as a bearer (see [Authentication & Agent Credentials](/connect/authentication.md)):

  ```bash
  export MEDIAFIER_TOKEN=<MEDIAFIER_TOKEN>
  ```

Whichever path you use, your organization context is derived from your identity server-side. You never set a tenant header.

***

## 3. Make your first call

Start by listing the tools available to your organization — a read-only call that's always safe to run. In an agent, just ask it what Mediafier tools it can see. Over raw MCP, it's one JSON-RPC request:

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/list"
}
```

Once you can see the available tools, calling one follows the same shape — pick a tool name from the list and pass its arguments with `tools/call`. Every call runs through one enforcement chain (identity → organization → authorization → rate limit → billing → audit) before it executes, and every response carries a trace identifier for support correlation.

### Discover partner commands

Some partners publish **headline commands** — stable, agent-facing aliases over a governed tool with arguments pre-filled. To list the commands available to your organization, call the `list_commands` tool with no arguments:

```json
{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "list_commands",
    "arguments": {}
  }
}
```

Invoke a command by its `alias` exactly like any other tool — the gateway merges the command's fixed arguments and dispatches to the underlying tool, so a command call is audited identically to the equivalent raw call.

***

## 4. Next steps

* [MCP Access](/connect/mcp-access.md) — connect a specific client and discover everything available across the platform.
* [Authentication & Agent Credentials](/connect/authentication.md) — interactive sign-in and headless credentials, scopes, and rotation.
* [Troubleshooting](/connect/troubleshooting.md) — what an error means and what to do.