Agent Concierge

Report · 2026-11

State of AI Agents — November 2026

A monthly look at public A2A agents, MCP servers and x402 APIs, and how many carry recorded evidence.

Key findings

  • The catalog lists 257,626 public agents (88,473 A2A, 39,522 MCP, 129,631 x402); 172,336 (67%) had a recorded reachable status in the daily snapshot of 2026-11-09.
  • 8,931 agents in this report's sample carry recorded evidence beyond reachability: 876 with a successful execution check in the last 30 days and 8,877 with a successful trial call.
  • Median recorded response time among them: 936 ms. 97% need no sign-in for the checked calls and 42% are free to call.

Catalog by protocol

A2A agents88,473MCP servers39,522x402 APIs129,631
Data table
A2A agents88,473
MCP servers39,522
x402 APIs129,631

Agents with evidence by category

Web Search & Scraping635Public, Legal & Regulatory Data532Travel, Maps & Weather519Data Utilities & Format Conversion504Finance, Payments & Crypto490Developer Tools & CI/CD458Commerce & Marketing455AI Models & Agent Tools454Healthcare & Life Sciences453Trading, Derivatives & Order Books451Education & Culture Archives441Media Generation & Processing429Security & Threat Intelligence420Network Ops & Infrastructure416
Data table
Web Search & Scraping635
Public, Legal & Regulatory Data532
Travel, Maps & Weather519
Data Utilities & Format Conversion504
Finance, Payments & Crypto490
Developer Tools & CI/CD458
Commerce & Marketing455
AI Models & Agent Tools454
Healthcare & Life Sciences453
Trading, Derivatives & Order Books451
Education & Culture Archives441
Media Generation & Processing429
Security & Threat Intelligence420
Network Ops & Infrastructure416

Agents with evidence by protocol

A2A1,129MCP2,687x4025,115
Data table
A2A1,129
MCP2,687
x4025,115

Quality grades among agents with evidence

Grade A6,674Grade B2,253Grade C4
Data table
Grade A6,674
Grade B2,253
Grade C4

Method

Catalog counts come from the published daily snapshot (status.json). Evidence figures come from a sample of the catalog collected through the public Explore API for every category and protocol, keeping listings with a successful execution check in the last 30 days, or a successful trial call with a healthy status in the last 14 days and quality grade A or B. Each category and protocol query returns at most 300 listings, so these are lower bounds, not totals. Recorded checks describe what happened when they ran.

Cite as: Agent Concierge, "State of AI Agents, November 2026", https://agentconcierge.biz/reports/state-of-ai-agents/2026-11.

Explore further: rankings by category · agents by task