moda-mcp

Moda is an AI design agent. The MCP server at https://mcp.moda.app/mcp lets you create designs from a prompt, remix existing canvases, and export to PNG / JPEG / PDF / PPTX. Public share links (moda.app/s/…) work unauthenticated; everything else requires OAuth (default in Claude / Cursor / VS Code / Gemini CLI) or an API key (moda_live_… for scripts, CI, scheduled jobs).

Call whoami first

On any new conversation, call whoami before doing anything else. It returns identity + active workspace + entitlements in one payload (~no UI, no side effects) and lets you skip a lot of downstream tool calls:

• org_count == 1 → the user has one organization; no need to call list_organizations or ask which workspace to use.
• brand_kit_count == 1 → use team_default_brand_kit.id directly; don't call find_brand_kits.
• brand_kit_count == 0 → no kit exists; offer create_brand_kit(url=…) or proceed with skip_brand_kit=true.
• concurrency_cap → the bulk fan-out window for this plan (3 free / 10 paid / 15 ultra/enterprise). Use it instead of defaulting to 3.
• session.brand_kit_id → if non-null, the user has pinned a kit for this session; honor it.

If whoami shows org_count > 1 and the active session isn't set, call set_context(org_name=…) (or ask the user) before any workspace-scoped tool.

Pick a design-task recipe

After whoami, the next thing to do is match the user's intent to one of five canonical workflows. Open the matching recipe before calling start_design_task — each one covers the right parameters, what's preserved vs mutated, and what to surface back to the user.