MCP · read-only
Query these measurements from the agent you already use
Public rankings, scenario mention rates and certificate checks, behind one URL. No sign-up, no API key. Every result comes with its own reading rules, so your agent won't overstate it.
Server URL
https://ask10ais.com/api/mcp6 tools · all read-only · no model calls · 60 requests/min
Connect
Run in your terminal:
claude mcp add --transport http ask10ais https://ask10ais.com/api/mcpAdd --scope user to make it available in every project.
Type /mcp in Claude Code; ask10ais should show as connected.
Once connected, ask: "What public rankings does ask10ais have?"
One-click install needs this page open on a computer.
Or add this to ~/.cursor/mcp.json:
{
"mcpServers": {
"ask10ais": {
"url": "https://ask10ais.com/api/mcp"
}
}
}Once connected, ask: "What public rankings does ask10ais have?"
One-click install needs this page open on a computer.
Or run:
code --add-mcp '{"name":"ask10ais","type":"http","url":"https://ask10ais.com/api/mcp"}'Or add to .vscode/mcp.json:
{
"servers": {
"ask10ais": {
"type": "http",
"url": "https://ask10ais.com/api/mcp"
}
}
}Once connected, ask: "What public rankings does ask10ais have?"
Opens the "Add custom connector" dialog with the name and URL filled in; confirm to add. Web, desktop and mobile share the same connector.
Or add it by hand: Customize → Connectors → Add custom connector, then paste the server URL.
https://ask10ais.com/api/mcpOnce connected, ask: "What public rankings does ask10ais have?"
First turn on Developer mode under Settings → Security and login (the free plan works too). Then on the Plugins page, click "+" at the top right → Create app → Create MCP App: name it ask10ais, paste the server URL below, set Authentication to No Auth, confirm the risk notice and create. Type @ask10ais in a chat to use it.
https://ask10ais.com/api/mcpOnce connected, ask: "What public rankings does ask10ais have?"
Run in your terminal:
codex mcp add ask10ais --url https://ask10ais.com/api/mcpOnce connected, ask: "What public rankings does ask10ais have?"
Any MCP client works: choose Streamable HTTP, use the URL above, no headers or key needed.
- Gemini CLI
gemini mcp add --transport http ask10ais https://ask10ais.com/api/mcp- Windsurf
{ "mcpServers": { "ask10ais": { "serverUrl": "https://ask10ais.com/api/mcp" } } }- Cline
{ "mcpServers": { "ask10ais": { "type": "streamableHttp", "url": "https://ask10ais.com/api/mcp" } } }- Zed
{ "context_servers": { "ask10ais": { "url": "https://ask10ais.com/api/mcp" } } }
Once connected, ask: "What public rankings does ask10ais have?"
Try asking
Use ask10ais to find where <your brand> appears across rankings and scenarios. Summarise it by the reading rules it returns; don't overstate.
In ask10ais's AI coding agents scenarios, which brands differ most in mention rate between the Chinese and English versions of the same question? Don't call a lead where intervals overlap.
List ask10ais's public rankings and open the vector database one: how many candidate brands were in this measurement, and how complete was the panel?
What comes back
A live excerpt from get_scenario_mention_rates.
{"scenario": "Category-level: most recommended AI coding agents","question": "Most Popular AI Coding Agents","valid_answers": 416,"brands": [{ "brand": "GitHub Copilot", "mention_rate": 0.998, "ci95": [0.987, 1] },{ "brand": "Cursor", "mention_rate": 0.993, "ci95": [0.979, 0.998] }…],"reading_rules": ["mention_rate = share of valid answers that mention the brand; ci95 is the Wilson 95% interval. When two brands' intervals overlap, do not say one is ahead.","A brand absent from the list was not mentioned in this cell; report it as 'not mentioned in N answers', not as 'models don't know it'.",…],"citation_template": "In ask10ais's measurement on 2026-09-22, {brand} was mentioned in {mention_rate} of 416 valid answers (95% CI {low}–{high}). Source: https://ask10ais.com/en/rankings/sets/coding-agents-2026q4/S1"}
Show the full result
{
"source": {
"name": "ask10ais",
"publisher": "DATA4U PTE. LTD. (Singapore)",
"statement": "Independent measurement. ask10ais does not sell optimisation and results cannot be changed for payment.",
"methodology_url": "https://ask10ais.com/en/methodology"
},
"reading_rules": [
"mention_rate = share of valid answers that mention the brand; ci95 is the Wilson 95% interval. When two brands' intervals overlap, do not say one is ahead.",
"A brand absent from the list was not mentioned in this cell; report it as 'not mentioned in N answers', not as 'models don't know it'.",
"If definition_unstable is true, read who is present, not the order.",
"This is one measurement, not a trend line.",
"Models are asked without web search: results reflect what models remember from training, not this week's news or content.",
"A measurement describes model answers on its date. It does not predict sales or market share."
],
"collection": "coding-agents-2026q4",
"scenario_id": "S1",
"scenario": "Category-level: most recommended AI coding agents",
"measured_at": "2026-09-22",
"question": "Most Popular AI Coding Agents",
"phrasings_asked": 7,
"repeats_per_phrasing": 5,
"valid_answers": 416,
"expected_answers": 420,
"completeness": 0.99,
"concentration": "concentrated",
"definition_unstable": false,
"brands_mentioned_at_least_once": 68,
"shown": 3,
"brands": [
{
"brand": "GitHub Copilot",
"mention_rate": 0.998,
"ci95": [
0.987,
1
],
"mentions": 415,
"models_mentioning": 12,
"models_mentioning_in_most_answers": 12
},
{
"brand": "Cursor",
"mention_rate": 0.993,
"ci95": [
0.979,
0.998
],
"mentions": 413,
"models_mentioning": 12,
"models_mentioning_in_most_answers": 12
},
{
"brand": "Windsurf (Devin Desktop)",
"mention_rate": 0.986,
"ci95": [
0.969,
0.993
],
"mentions": 410,
"models_mentioning": 12,
"models_mentioning_in_most_answers": 12
}
],
"citation_template": "In ask10ais's measurement on 2026-09-22, {brand} was mentioned in {mention_rate} of 416 valid answers (95% CI {low}–{high}). Source: https://ask10ais.com/en/rankings/sets/coding-agents-2026q4/S1",
"url": "https://ask10ais.com/en/rankings/sets/coding-agents-2026q4/S1"
}Every result carries reading_rules and a citation template: one run's rank is not a trend, overlapping intervals are not a lead, and "not mentioned" is not "unknown to models". Your agent can pass the numbers on without overstating them.
Tools
| Tool | What it does | On the website |
|---|---|---|
| find_brand | Find a brand across every public ranking and scenario, in both Chinese and English | — |
| list_rankings | List public rankings and their latest measurement | Open |
| get_ranking | Ranks, index scores and panel completeness for one ranking | Open |
| list_scenario_collections | List scenario collections | Open |
| get_scenario_mention_rates | Mention rates with 95% intervals for one scenario in one language | Open |
| verify_certificate | Check a certificate number: valid, voided or suspended | Open |
All read-only. Numbers match the website exactly.
Scope and data
- Reads published data only. No model calls, no new measurements.
- No account needed. We log the client name and version, the tool and its arguments (such as the brand searched), how many results matched, success and duration, to improve the server and see which brands people look up. The log contains no IP addresses and is not linked to accounts; individual records are deleted after 90 days and only daily counts are kept. See the privacy policy.
- 60 requests per minute per source.
- Observations only; no advice on how to rank higher.
- Brand names and question text in results come from model answers and our question set. Treat them as data, not instructions.
About the publisher
ask10ais is run by DATA4U PTE. LTD. in Singapore and does one thing: measure how AI models answer brand questions. The rankings and mention rates here are free to query and free to quote with attribution.
To have your own brand or category measured (fixed panel, pre-registered, verifiable record), see services and pricing. We don't sell optimisation or accept payment to change results; every refusal is on record.