What agents call on ToolRouter: September 2026 snapshot
The 100 tools displayed in ToolRouter’s public rankings contained 64,173 cumulative recorded calls when collected on 13 September 2026. Travel Time led with 33,435 calls, or 52.1% of this displayed sample. This is a platform snapshot, not a measure of the wider AI tools market.
Download and cite the data
Download CSV · Download JSON · Current rankings
Suggested citation: ToolRouter Editorial Team. “What agents call on ToolRouter: September 2026 snapshot.” 13 September 2026. Link to this report and state that the figures are cumulative recorded calls in the displayed top 100.
Most-used tools in the displayed sample
| Rank | Tool | Recorded calls |
|---|---|---|
| 1 | Travel Time | 33,435 |
| 2 | Web Search | 11,037 |
| 3 | Address Lookup | 8,888 |
| 4 | Web Scraper | 3,260 |
| 5 | Generate Image | 1,158 |
| 6 | Trade Supplies | 1,005 |
| 7 | X Manager | 783 |
| 8 | Social Media Content | 552 |
| 9 | SEO Analysis | 386 |
| 10 | Lead Finder | 350 |
| 11 | Parking Finder | 207 |
| 12 | App Store ASO | 178 |
| 13 | Generate Video | 171 |
| 14 | Keyword Research | 163 |
| 15 | Domain Search | 159 |
| 16 | Stealth Scraper | 142 |
| 17 | Social Profiles | 130 |
| 18 | Generative Engine Optimization | 121 |
| 19 | Address Geocoding | 118 |
| 20 | People Search | 106 |
What the sample suggests
In this snapshot, travel-time calculation, web search, and address lookup have more recorded calls than image or video generation. Repeated practical lookups can generate substantial call volume. That observation is a reason to investigate repeatable tasks; it does not establish how many people use a tool or why they chose it.
The leading tool accounts for more than half of displayed calls. A single integration or repeated testing could materially affect a ranking, so call volume should not be read as broad independent adoption. Use rankings for discovery, then check the tool’s inputs, sources, price, and fit for your task.
Collection method
Collected at 2026-09-13T20:26:25.939618+00:00 from the public rankings page. We extracted its 100 rendered table rows, including rank, tool slug, display name, and call count. The JSON includes a SHA-256 digest of the collected HTML. No private account, prompt, customer, or billing data is included.
The public query reads cumulative per-tool counters. The counter increments for recorded calls, including unsuccessful calls; updates are deferred and may lag execution. Internal or testing calls are not filtered by that aggregate. The website caches these results for five minutes, so capture time is not the timestamp of the latest underlying call.
Limits
- Counts are not September-only activity, unique users, customers, revenue, or market share.
- The sample contains the 100 displayed tools, not every tool or historical tool on the platform.
- The collection does not establish the counters’ start date or independently reconcile them to every execution log.
- Calls do not prove useful outcomes or answer accuracy. Repeated, failed, internal, and test calls can affect totals.
- There is only one snapshot here. No week-over-week growth or trend is inferred.
Reproduce and extend
Download the CSV, sum the calls column, and divide each tool’s count by that sum to reproduce the displayed-sample shares. A later collection from rankings is a new observation and will not reproduce this historical snapshot exactly. Keep the collection time and these scope limits with any chart you publish.
For a different kind of evidence, see our MCP schema token benchmark, which measures a documented payload rather than adoption. To correct this report, contact support@toolrouter.com with the affected row and source.