Skip to content
agent-media docs
Public documentation · No sign-in required
Documentation / Connect

One URL, then sign in.

Choose your client, add the hosted connector, and sign in to your Agent Media account. Then check your connection before generating.

On this page

The connector

Streamable HTTP
https://api.agent-media.ai/mcp

Per client

Claude.ai / Claude Desktop / Cowork

Customize > Connectors > + > Add custom connector > paste the URL > Connect and sign in. Enable Agent Media in the conversation connector menu.

Claude Code

claude mcp add --transport http --scope user agent-media https://api.agent-media.ai/mcp

Available in all your projects. In Claude Code, run /mcp, select agent-media, and complete browser sign-in. If already connected, do not add a second copy.

ChatGPT (web)

Enable Developer mode where your account or workspace permits it. Create a custom MCP app with the connector URL and OAuth, complete sign-in, then select Agent Media in the conversation. Refresh the app tools when the catalog changes.

Settings labels and availability vary by account and workspace policy. See the linked official ChatGPT setup guide below.

Codex

codex mcp add agent-media --url https://api.agent-media.ai/mcp

Then run codex mcp login agent-media to complete browser sign-in. Keep your normal approval settings; review requested tool actions.

Grok CLI

grok mcp add --transport http agent-media https://api.agent-media.ai/mcp

Cursor

{ "mcpServers": { "agent-media": { "url": "https://api.agent-media.ai/mcp" } } }

~/.cursor/mcp.json

Any stdio client

npx -y -p @agentmedia/mcp-server@latest agent-media-mcp

A stdio proxy to the same hosted server; set AGENT_MEDIA_API_KEY. Do not register both the connector and the proxy in one client, or every tool appears twice.

Official ChatGPT setup and tool refresh guide ↗

Paste to your agent

Use this with an agent that can configure MCP connections. Complete any browser sign-in yourself; otherwise follow the client instructions above.

Set up agent-media so I can generate videos, images and voice from here.
1. Add the MCP server: https://api.agent-media.ai/mcp (Streamable HTTP).
2. Authenticate: complete the sign-in in the browser it opens.
3. Call list_models and tell me what you can make.
4. Open the existing image upload panel using open_upload_panel, or upload_image with {} if only the older tools are available. Do not generate anything yet.

Sign in

OAuth 2.1 with dynamic client registration (browser sign-in); `Authorization: Bearer ma_...` API keys also accepted.

Connection checks and uploads use no generation credits. Creating media requires credits on the account. 1 credit = $0.01. Sign up at agent-media.ai; an API key comes from agent-media login or the dashboard.

Check setup before generating

Start a new conversation with Agent Media enabled and ask:

Use Agent Media to check my account balance, list available models and open its image upload panel. Do not generate anything yet.

Call get_account for the authenticated balance; list_models includes the same check for older connector catalogs. Opening the panel checks upload access separately. Compare the intended generation quote with your balance.

If your agent cannot see the new panel tools, ask it to call upload_image with an empty object. It returns the same uploader and instructions for retrieving your images. Images expire 24 hours after the upload session opens.

Polling

Every generate tool returns a job id. Call get_run_status with wait:true; each call blocks up to about 45 seconds, a video usually needs several calls (seedance-2.0: about 3 minutes for a 5 s clip at 720p; seedance-2.5: 12 to 25 minutes per clip; plan the wait). Never report a result before get_run_status returned its URL.