Mintlify

@mintlify show me the top 10 search queries from the last 30 days and their click-through rates

Here are the top searches from your docs in the past month: 1. "authentication setup" — 847 searches, 68% CTR, top page: /guides/auth 2. "API rate limits" — 612 searches, 71% CTR, top page: /api-reference/limits 3. "webhook configuration" — 489 searches, 54% CTR, top page: /webhooks/setup 4. "error codes" — 423 searches, 62% CTR, top page: /troubleshooting/errors 5. "SDK installation" — 391 searches, 77% CTR, top page: /sdks/getting-started (5 more entries follow) This data reflects what users are actively looking for. The lower CTR on "webhook configuration" suggests that page may need clearer headings or better search optimization.

This example uses Get Search Queries to surface content gaps. The MCP returns paginated results ordered by search volume; you'll need to specify a date range. No destructive side-effects. Useful for identifying documentation blind spots before investing in agent-driven rewrites.

@mintlify create an agent job to add a troubleshooting section to the authentication guide covering common OAuth errors

Agent job created successfully (ID: job_8x3k9m2p). The documentation agent is now analyzing your authentication guide and drafting a troubleshooting section for common OAuth errors. Status: running Target repo: your-org/docs Branch: main Model: gpt-4 The agent will open a pull request when complete. You can review changes before merging. Estimated completion: 3-5 minutes. I'll poll the job status and notify you when the PR is ready.

This example demonstrates Create Agent Job, which is DESTRUCTIVE and consumes Mintlify credits. The agent autonomously edits your docs repository and opens a PR — always review before merging. Requires write access to the connected repo. Use Get Agent Job to poll for completion and retrieve the PR link.

@mintlify compare AI assistant usage versus human page views for our API reference section over the last quarter

Here's the traffic breakdown for /api-reference/* from the past 90 days: **Human visitors:** 12,847 page views across 23 pages - Top page: /api-reference/authentication (2,103 views) - Average time on page: 2m 14s **AI assistant queries:** 8,492 questions answered - 71% resolution rate (user marked as helpful) - Most cited pages: /api-reference/rate-limits, /api-reference/webhooks **Insight:** AI traffic represents 40% of total API reference engagement. The assistant is heavily citing rate-limit docs, but human visitors spend more time on authentication — consider surfacing rate-limit context earlier in the auth flow.

This example combines Get Page Views (which splits human vs AI traffic) with Get Assistant Conversations to synthesize usage patterns. Requires date-range filtering on both endpoints. The MCP returns raw data; the AI performs the comparative analysis and surfaces actionable recommendations.

When the agent job beats hiring a tech writer for launch week

A 6-person startup shipping a new API endpoint in 5 days needs reference docs written and merged before Friday. The Create Agent Job tool takes a natural-language prompt ("document the /v2/webhooks endpoint with code samples in Python and Node"), edits the connected repo, and opens a PR—all in one background job. You review the diff, request changes in plain English via a second agent job, and merge. This works when your docs repo is already Mintlify-structured and the endpoint surface is under 20 methods; beyond that, the agent starts hallucinating parameter types. If you're launching a 60-endpoint platform rewrite, hire the writer. For a 3-endpoint feature drop, the agent job closes the docs gap in an afternoon and your API key covers the credits.

Why search query export matters more than conversation logs

A 12-person SaaS support team sees the same integration questions every week but can't pinpoint which doc pages are failing. The Get Search Queries tool exports the top 50 search terms with click-through rates and the pages users land on after searching. You discover "webhook retry logic" has 340 searches and a 12% CTR—people aren't finding the answer. You rewrite that section, ship it, and watch CTR climb to 68% over two weeks. The Get Assistant Conversations tool gives you the raw Q&A transcripts, but search queries tell you what's missing before users even ask the AI. If your docs get under 500 searches a month, the signal is too noisy. Above that threshold, this becomes your quarterly content roadmap.

When feedback-by-page data drives the rewrite backlog

A 20-person product org reviews docs health every quarter and needs to rank pages by user dissatisfaction. The Get Feedback By Page tool returns thumbs-up/down counts per URL, so you sort by negative feedback density and surface the 10 pages with the worst ratios. You cross-reference Get Page Views to confirm those pages also have high traffic (a page with 3 downvotes and 8 views isn't the priority; one with 40 downvotes and 2,000 views is). This two-tool combo builds the rewrite backlog in 20 minutes. The Get Feedback tool gives you the raw comment text, but that's a second pass—you triage by aggregate scores first. If your docs get under 100 feedback events per quarter, the sample size is too small to trust the ratios.