agentrank — What agents call: usage from the platforms that run the tools — openbash

Calls and calling accounts per tool on Monid, MCP server use on Smithery, users and runs per tool on Apify, MCP image pulls on Docker, and which tools AI assistants recommend.

agentrank

agentrank is a command on the openbash shell, reachable over HTTP, SSH and MCP with one key. It composes with the other programs through pipes; there is no per-program endpoint and no schema for the join.

agentrank apify 25 --category SOCIAL_MEDIA

help

agentrank: agent tool usage, as the platforms that run or route it count it.

WHAT IT IS
  A download counts an install and a visit counts a person. Neither counts an
  agent at work. The platforms that run or route tools do:
    monid     calls and calling accounts per provider and endpoint
    smithery  the use count of each MCP server it lists or hosts
    apify     users (7 and 30 days, total) and runs (30 days) per store tool
    docker    all-time pulls of each MCP server image in Docker's catalog
    answers   asks ChatGPT, Perplexity, Gemini, AI Mode and Copilot a question
    answer    reads one assistant's answer, and which of your tools it names first

ASK, THEN READ
  An assistant takes 20 to 60 seconds and they answer one at a time, so
  `answers` returns at once with a task id per assistant. Read each with
  `agentrank answer <task> --tools A,B,C`: QUEUED or PROCESSING means wait and
  read again; COMPLETED carries the answer. A task is kept for 24 hours.

ANSWERS ARE READINGS, AND COST
  Each assistant asked is one agentrank_answer on the bill; reading is free.
  An answer is what one assistant said once: ask more than once, or several
  questions, before calling anything a pattern. The other studies are free.

ONE PLATFORM, NEVER A MARKET SHARE
  Each counts only its own traffic. Monid sees only calls routed through it; a
  provider whose customers call its own API directly reads SMALL there however
  big it is. Smithery counts its servers, Apify runs on Apify, Docker pulls of
  its images (installs, not calls). Say which platform a figure is from, and
  compare platforms side by side, never by adding them up.

ACCOUNTS BEFORE CALLS
  A call counts once whatever it does, so a search an agent runs thirty times
  a task outranks a render it runs once. `accounts` is how many accounts call an
  endpoint; 60,000 calls from 62 accounts is a few pipelines, not adoption.
  A provider's `accounts_at_least` is its busiest endpoint's figure — a floor.

WEIGHTED, AND MOSTLY FREE
  Monid ranks by a weighted count: a free call is a fifth of a paid one. The
  headline gives raw and paid calls, and free_share — most traffic is free.
  `change` is against the window before, while the platform itself grows; a
  provider growing slower than the headline is losing share.

Studies:
  monid    <view>
      calls, calling accounts and change per provider or endpoint, with the free share of all traffic
  smithery <query>
      MCP servers matching a name, with Smithery's use count for each
  apify    <limit>
      Apify Store tools with users in the last 7 and 30 days and runs in the last 30 days
  docker   <limit>
      MCP server images in Docker's MCP Catalog by all-time pulls
  answers  <prompt>
      ask ChatGPT, Perplexity, Gemini, AI Mode and Copilot a question; answers come back as task ids to read with `agentrank answer` — billed per assistant
  answer   <task>
      one assistant's answer once it is done, its sources, and which of your tools it names first

JSON to stdout, so it composes: pipe to `jq` to pick a field, or `> /tmp/…` to keep it.
Run a study with no arguments — `agentrank monid` — to see what it takes.

What it answers

agentrank monid — calls, calling accounts and change per provider or endpoint, with the free share of all traffic
agentrank smithery — MCP servers matching a name, with Smithery's use count for each
agentrank apify — Apify Store tools with users in the last 7 and 30 days and runs in the last 30 days
agentrank docker — MCP server images in Docker's MCP Catalog by all-time pulls
agentrank answers — ask ChatGPT, Perplexity, Gemini, AI Mode and Copilot a question; answers come back as task ids to read with `agentrank answer` — billed per assistant
agentrank answer — one assistant's answer once it is done, its sources, and which of your tools it names first