AsimovBox Developers — REST API, CLI, MCP & A2A
AsimovBox
Build with AsimovBox

Long-form video generation for your own workflow.

Use the studio when you want visual control, or connect AsimovBox to an agent, script, or internal tool through REST, the Python CLI, hosted MCP, or A2A.

Talk to support

REST API

Create videos, inspect status, update scenes, and render through authenticated JSON endpoints.

POST /api/v1/videos

Python CLI

Use the AsimovBox package from scripts, notebooks, CI jobs, or internal tools.

pip install asimovbox

Hosted MCP

Connect Claude and other OAuth-capable agents to the Streamable HTTP video server.

https://asimovbox.com/api/video-mcp

A2A video agent

Let another agent delegate complete video tasks and retrieve status, cancellation, and MP4 artifacts.

https://asimovbox.com/.well-known/agent-card.json

OpenAI plugin

Use one OAuth connection and guided video workflow across supported ChatGPT and Codex surfaces.

AsimovBox Video · universal plugin

DeepSeek Harness

Install native Harness tools backed by the same hosted MCP and production rendering pipeline.

dsh plugin --profile default add github:cerebrixos-org/dsh-asimovbox

Hosted MCP setup

Use the hosted endpoint when your client supports OAuth and Streamable HTTP. The client discovers authorization metadata, opens the AsimovBox consent flow, and receives a scoped bearer token. No local package or process is required.

  1. Add https://asimovbox.com/api/video-mcp as the MCP server or connector.
  2. Complete the OAuth sign-in and consent flow.
  3. Use the video tools to create, update, render, and inspect videos.

For clients that only support local stdio servers, use the Python CLI package and configure the user API key from Account → API access.

A2A setup

Point an A2A v1 client at the public Agent Card. A new user can connect directly from the agent: AsimovBox opens Google sign-in and consent in the browser, creates the account when needed, and returns a scoped token to the agent. The user does not need to visit Studio or copy an API key.

  1. Discover https://asimovbox.com/.well-known/agent-card.json and dynamically register the OAuth client.
  2. Complete Authorization Code + PKCE using the advertised endpoints and the https://asimovbox.com/api/a2a resource.
  3. Send SendMessage, GetTask, ListTasks, or CancelTask to the advertised interface with A2A-Version: 1.0.
  4. Use a structured application/json part to explicitly request paid stages such as generate_storyline, render, or finalize.

A personal API key remains available for existing server-to-server integrations. A plain-text request creates a free draft only; AsimovBox will not infer authorization to spend credits from ambiguous prose.

DeepSeek Harness setup

Install the native Harness bundle, generate a personal API key under Account → API access, and provide it through the environment. The plugin registers the same ten discovery and video workflow tools exposed by hosted MCP.

  1. Run dsh plugin --profile default add github:cerebrixos-org/dsh-asimovbox.
  2. Set ASIMOVBOX_API_KEY without committing it to source control.
  3. Start Harness with the profile and ask it to compare video options or create a free draft.

Storyline generation, rendering, and branded finishing can consume credits. The plugin distinguishes free draft creation from those paid stages and requires explicit user intent.

OpenAI plugin

The AsimovBox Video package combines the hosted OAuth MCP server with a guided workflow skill. It is designed for the universal plugin directory shared by supported ChatGPT and Codex surfaces, while direct MCP connections remain available for developer-mode and private use.

Designed for agents and search

The product workflow is intentionally structured: brief → storyline → editable scenes → voice → render. The public llms.txt file summarizes the product, endpoints, and canonical pages for crawlers and AI assistants.