Quickstart

Point any MCP client at the keyless trial endpoint and make a real call — no account, no human in the loop. Every stateless tool, same contracts, 25 calls a day per network.

$ claude mcp add openaso-trial --transport http https://mcp.openaso.ai/trial

Tracking, history, webhooks and the App Store Connect tools need an account. The budget error tells your agent exactly that.

Create an account, then mint an API key in your dashboard. Clients that sign in with OAuth (Claude.ai and ChatGPT connectors, VS Code) can skip the key: add https://mcp.openaso.ai/mcp, approve once in the browser, done — the token maps to the same account.

The key is shown only once — the dashboard also shows usage and rotates or revokes keys. Free while in early access: 500 tool calls per day and 500 tracked keyword × app × storefront combinations.

$ claude mcp add openaso --transport http https://mcp.openaso.ai/mcp \
    --header "Authorization: Bearer YOUR_KEY"

Cursor, Windsurf, or anything MCP — same endpoint, same header:

{
  "mcpServers": {
    "openaso": {
      "type": "http",
      "url": "https://mcp.openaso.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_KEY" }
    }
  }
}

x-api-key: YOUR_KEY works too if your client cannot set an Authorization header. The server is stateless Streamable HTTP — POST only, no session juggling.

Good first prompts, in rough order of ambition:

• "Where does Streaks rank for habit tracker in the US and UK?" — find_app resolves the name to a trackId, then one rank call.
• "Audit the App Store listing for <your app> in the US and propose better metadata" — the audit tool gathers listing, word pool, and scored keyword candidates in one call, then your agent drafts against coverage.
• "Track our top 20 keywords daily and tell me when anything drops" — track_config + the managed daily snapshot run, then history dropsOnly=true any morning, or a webhook so the drops come to you.
• "Read our real keyword field from App Store Connect, tighten it, and ship it" — asc_connect once with an ASC API key, then asc_listing → coverage → asc_update_listing, and annotate the change so next week's history reads as before/after.
• "Did the new listing convert better?" — asc_analytics_enable once, then asc_analytics for the real impressions → page views → downloads funnel per storefront and source, set against history for the same days.

Before a long run, account reports the remaining daily budget and tracked capacity — agents should check it the way they would check disk space.

There is also a packaged optimize_metadata prompt on the server — the full playbook from audit to tracked verification. More worked prompts — audits, keyword research, competitor maps, daily tracking — live in the recipes.