What you can do
- Generate ad copy grounded in your brand kit (
generate_ad_copy) - Query performance across campaigns, ad sets, and ads (
query_performance) - Discover models — image, video, TTS, and lipsync registries with cost and capability metadata (
list_models) - Browse avatars and voices for talking-actor videos (
list_avatars,list_voices) - Manage org data — team info, credit balance, brand kit, products, projects, ad accounts
- Search your asset library and inspect campaigns and launches
Multi-model by design
AdStellar is multi-model. The MCP does not expose one tool per model — instead:- Call
list_modelswith acapabilityfilter (image,video,tts,lipsync) to enumerate available models, their providers, costs, and limits. - Pass the chosen model’s
idas themodelparameter of a generation tool. Omit it to use the default router (best quality-per-credit for the job).
Credits & billing
Image and video generation over the MCP server and the REST API draw AI credits from your team’s balance, exactly like generating inside the app. Ad copy and all read-only requests (analytics, model/avatar listings, status checks, org data) are free. Check your balance anytime with the
get_credit_balance tool, and see per-model costs via list_models.Authentication
The MCP uses team-scoped API keys (ast_live_...). See Authentication. OAuth 2.1 is on the roadmap.
Next steps
Authentication
Create and scope your API key.
Quickstart
Connect Claude Desktop in two minutes.
Tool reference
Every tool, its scopes, and its parameters.