AgentMediaAgentMedia

Install · agent-media

Generate AI UGC videos from your agent

agent-media ships as a CLI, an MCP server, and a Claude Code skill — pick whichever fits your client. All paths talk to the same API and bill from the same balance.

github.com/gitroomhq/agent-medianpmagent-media-cli

Pick your client

The install path is different per client. Find yours below.

Claude Code

Anthropic's terminal CLI

# 1. Install the CLI (1.12.0+) + log in
npm install -g agent-media-cli@latest
agent-media login

# 2. Add the skill — pulls SKILL.md + 10 reference files
claude skills add github:gitroomhq/agent-media

# 3. (later) keep the skill fresh
agent-media skill update

Restart Claude Code. The skill asks 4 quick questions (topic, who, duration, cost) then drives agent-media.

Cursor / Continue / Windsurf

MCP-aware IDEs

# 1. Install the CLI
npm install -g agent-media-cli@latest
agent-media login

# 2. Add the MCP server to your client's mcp.json:
{
  "mcpServers": {
    "agent-media": {
      "command": "npx",
      "args": ["-y", "@agentmedia/mcp-server"],
      "env": { "AGENT_MEDIA_API_KEY": "<your ma_xxx>" }
    }
  }
}

Reload the IDE. Tools create_selfie, create_character, create_subtitle show up.

Claude.ai / Cowork / Claude Desktop

Web app + desktop app — hosted HTTP MCP

Settings → Integrations → Add custom MCP server:

URL:    https://api.agent-media.ai/mcp
Auth:   Bearer ma_xxx   (your agent-media API key)

Tools create_selfie, create_character, create_subtitle show up. Don’t have an API key? Run agent-media login on any machine first.

Standalone CLI / scripts / CI

No agent — just the CLI

npm install -g agent-media-cli@latest
agent-media login
agent-media selfie --help

Every command supports --json and --quiet for piping. Add --sync to block until the job completes.

Plain REST

From anything that can curl

curl -X POST https://api.agent-media.ai/v2/selfie \
  -H "Authorization: Bearer ma_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "description": "25yo asian woman, long wavy dark hair, soft smile",
    "script": "I keep getting DMs about my hair oil routine",
    "preset": "salon-mirror-result",
    "vibe": "sassy",
    "duration": 10
  }'

Poll GET /v1/videos/<job_id> until status: completed. Full reference in the repo.

What you get

agent-media selfie

9:16 TikTok-style video of an AI person talking to camera. 4-stage pipeline. No photo needed — generates the character from your description. Durations: 5, 10, or 15s.

agent-media character create

Persist a reusable AI character. Reuse the char_xxxxxxxxxx across every video for on-model consistency.

agent-media subs

Burn styled subtitles onto any existing video. 17 styles (Hormozi, neon, minimal, …). Whisper auto-transcribes.

Full API referencePricingSign in