Tool limits and performance
Every cap that applies when you run a tool: rate limits, crawl and row limits, timeouts, retention, and how long each kind of tool takes.
Last updated 2026-08-06
Summary#
This page collects every limit that can affect a tool run: how many calls you can make in an hour, how many pages a crawler will visit, how long a tool waits before it gives up, and how long your results are kept. Use it to tell an expected cap apart from a fault.
Overview#
Limits in Metric Vault fall into five groups.
| Group | What it protects | Where it bites |
|---|---|---|
| Rate limits | The platform, from runaway automation | 100 light-tool calls per hour |
| Quotas | Your plan's allowance | Premium tools, 3 credits and up |
| Work limits | Run time, by capping how much a tool inspects | Crawlers, row counts, sampling |
| Timeouts | You, from waiting forever | Long-running tools and workflows |
| Retention | Storage | Saved results, history, favorites |
Quotas are covered in Quotas and rate limits; everything else is here.
Rate limits#
| Limit | Value | Applies to | What you see |
|---|---|---|---|
| Light-tool fair use | 100 calls per clock hour | Every signed-in account | Hourly fair-use limit reached (100 light-tool calls/hour). This protects our infrastructure from abuse while keeping your monthly usage unlimited. Please wait ~<N> minutes for the next hourly reset. |
| Anonymous fair use | 100 calls per clock hour, per IP address | The public tools at metricvaultai.com/free-tools | The same message |
| Public API | Documented separately | API key holders | See Rate limits and quotas |
The hourly counter resets on the hour, in UTC, and the message tells you how many minutes remain. Only free and light runs consume it. Premium runs, meaning anything costing 3 credits or more, do not touch the hourly bucket; they are limited by your monthly allowance instead.
Note: A normal working session runs perhaps five to twenty light calls in an active hour. If you are hitting 100, something is looping, or a workflow is being re-run repeatedly.
Work limits, by tool#
These caps decide how much a tool looks at. They exist so a run finishes in seconds rather than minutes, and they are the reason a very large site is sampled rather than crawled exhaustively.
| Tool | Limit |
|---|---|
| Bulk Analysis | 10 domains per run |
| Duplicate Content Audit | Crawls up to 12 pages |
| Internal Linking Engine | Crawls up to 20 pages |
| llms.txt Builder | Lists up to 25 URLs from your sitemap |
| Plagiarism check | Samples 8 sentences from your text; needs at least 25 words |
| Fact Check | Extracts and verifies up to 6 checkable claims; needs at least 25 words |
| Localization | 100 strings and 20,000 characters per request |
| Growth Actions | Returns the top 20 ranked actions |
| Sitemap Validator | Follows nested sitemap indexes and reports a sample of URLs |
| Assistant message | 2,000 characters and 4 images |
| Public share link | Refuses to build from a result shorter than 60 characters |
Warning: The Bulk Analysis screen says "up to 20 domains". The tool actually processes the first 10 lines you paste. Split a longer list into two runs.
Several data tools ask their provider for a bounded number of rows. The most useful ones to know:
| Tool | Rows requested |
|---|---|
| Domain Overview | Up to 50 ranked keywords, 30 competitors, 10 top pages, 10 subdomains, 10 anchors, plus a 21-country breakdown |
| Organic Research | Up to 80 ranked keywords |
| Rank Tracker | Up to 100 ranked keywords |
| Content Decay Detector | Up to 700 ranked keywords |
| PPC Research | Up to 50 paid keywords and 10 paid competitors |
| Keyword Explorer | Up to 20 keyword suggestions per source |
| Backlink Profile | Up to 15 rows per supporting list |
| Link Prospector | Up to 6 competitors and 40 referring domains |
| Market Map | Up to 100 keyword suggestions and 12 competitors |
| Audience Overlap | Up to 3 competitor domains |
| Gap Finder | Up to 3 competitor domains |
Timeouts and run times#
| Tool or stage | Time budget | What happens when it is exceeded |
|---|---|---|
| Typical light tool | A few seconds | — |
| PageSpeed & Vitals | 75 seconds per attempt, 82 seconds overall | A stored report up to 7 days old is served instead of an error |
| Full Article Writer | 60 to 90 seconds typical | Runs as a background job you can poll; a job that has expired returns Job not found or expired. Please re-submit. |
| Auto-Execute, SEO Brief stage | 45 seconds | The pipeline stops; the failure card offers Retry full pipeline and Retry just the SEO Brief stage |
| Auto-Execute, Article Outline stage | 50 seconds | Falls back to a synthetic outline and continues |
| Auto-Execute, Article Writer stage | 110 seconds | The pipeline stops with a retry card |
| Auto-Execute, Content Optimizer stage | 60 seconds | Skipped if the writer returned nothing |
| Workflows | About 45 seconds typical, 30 seconds to 2 minutes | Individual steps that fail are reported in the consolidated report |
| Responsiveness analysis | 40 seconds for the AI pass | That site took too long to respond… |
Each Auto-Execute stage retries once, after a 1.5 second pause, if the failure looks transient: a timeout, a server error, or a network problem.
Tip: If a heavy tool times out on a very large site, re-run it. The second attempt often lands inside the cache window and returns immediately. See Result caching and freshness.
Retention#
| Item | Kept for |
|---|---|
| Saved results, including the rendered report | 90 days |
| A single saved result | Up to about 400 KB of data; larger runs are not stored |
| Activity history | See Report history |
| Favorite domains | 20, stored in your browser |
| Recent analyses | 12, stored in your browser |
| Recent sections in the command palette | 8, stored in your browser |
Favorites and recents live in the browser you used, so they do not follow you to another device. Saved results are stored in your workspace and do.
Team and plan limits#
| Limit | Free | Starter | Pro | Agency | Enterprise |
|---|---|---|---|---|---|
| Monthly credits | 0 | 100 | 500 | 2,000 | 10,000 |
| Invited teammates | 0 | 0 | 4 | 14 | Effectively unlimited |
The owner is not counted against the invite limit, so a Pro workspace holds five people in total. See Seats and plan limits and Plans and pricing.
What is not limited#
- The number of tools you can have open. Switching sections does not cancel a run.
- Free technical tools, beyond the hourly fair-use ceiling. They never consume monthly credits.
- Light tools on a paid plan. They keep working after your premium allowance is spent, which is why your usage counter can read higher than your quota.
- Reopening saved results, exporting, and sharing. None of these cost credits.
- Browser-only tools: image resize, image reformat, the live writing editor, the GEO scorer and the image auditor. They never call the server.
See also
Was this article helpful?
Thanks — feedback noted for the docs team.