agenthook

CLI for the AgentHook hosted media-generation API. Built for agents: progress goes to stderr, output URLs go to stdout, exit codes are meaningful.

npm i -g getagenthook        # or: npx getagenthook <command>

agenthook login --key <your-api-key>
agenthook tools
agenthook run make_video --prompt "A barista holds a latte and says: try our new oat flat white" --quality pro --aspect-ratio 9:16 --captions
agenthook run make_image --prompt "studio shot of a ceramic mug" --count 2
agenthook run caption_video --video-url https://…/video.mp4 --style chunk
agenthook list --search "flat white"
agenthook balance
agenthook history

• Credentials live at ~/.agenthook/credentials.json (chmod 600).
• API base: --api-url flag > AGENTHOOK_API_URL env > stored value > https://getagenthook.com; all requests hit <base>/api/v1/….
• Reference images (--ref, repeatable) require --owns-references: you attest you own, or have the rights to use, the likeness of every person appearing in the referenced images.
• Everything deterministically checkable is validated locally against the schemas served by GET /v1/tools (cached 1h) before any run is submitted: consent, prompt length caps, enum values, ranges, and nano-banana-2 (edit-only) without references.

MCP server

If your agent speaks MCP, it can call AgentHook directly instead of going through the CLI. The server ships as @getagenthook/mcp and runs over npx. Add this to your MCP client config (Claude Code, Cursor, and similar):

{
  "mcpServers": {
    "agenthook": {
      "command": "npx",
      "args": ["-y", "@getagenthook/mcp"],
      "env": { "AGENTHOOK_API_KEY": "ah_your_key" }
    }
  }
}

It exposes make_video, make_image, caption_video, and create_influencer, plus get_run. The generation tools return a run_id; pass it to get_run to fetch the finished URL once the job is done.

Develop

npm run build (emits dist/), npx vitest run, npx tsc --noEmit.