MCP (Model Context Protocol) is an open standard created by Anthropic that lets AI apps connect to external tools over a common interface. Think of it as “USB-C for AI tools”.
Endpoint
- Transport: HTTP (streamable)
- Auth: OAuth 2.1 with PKCE — no API keys
- Stateless: every request authenticates with a Bearer token
What you can do from your AI
Once connected, your AI has access to 6 tools:| Tool | Description |
|---|---|
search_tools | Find the right tool for a task (“generate image”, “text to speech”) |
get_tool_detail | Get the input schema and pricing for a specific tool |
generate | Submit a generation job (async — returns a job_id) |
check_generation | Poll a job until it completes |
list_generations | List your recent generations |
get_account | Check credits remaining and plan info |
Why MCP instead of the API?
| MCP | REST API | |
|---|---|---|
| Use case | Chat/editor workflows | Programmatic / production |
| Auth | OAuth (sign in once) | API key |
| Setup | 1-line install | Write code |
| Best for | Exploration, quick tasks | Apps, automation, batch jobs |
How it works
Authenticate
First request triggers an OAuth flow: a browser opens, you sign in to Artificial Studio, approve access. A token is stored locally.
Async generation — important
Image and especially video generation can take seconds to several minutes. The MCP server handles this with an async pattern:generatereturns immediately with ajob_idandstatus: "pending".- Your AI client polls
check_generationwith thatjob_idevery few seconds. - When
status: "success", theoutputfield contains the result URL.
poll_after_seconds as a hint so your AI waits the right amount of time. This prevents HTTP timeouts on long jobs (like Veo 3.1 videos).
See Async & polling for details.
Next steps
Quickstart
Install the MCP server in your AI editor — 60 seconds.
Authentication
How OAuth works and how tokens are stored.
Tool reference
Complete reference of all 6 MCP tools.
Async & polling
How long-running generations work under MCP.