AI agent leaderboard

The most used AI agents ranked by real usage: tokens routed through OpenRouter each day. Captured daily from the public apps board, kept here with history and in a crawlable form.

Hermes Agent | OpenRouter currently leads the board, processing about 814B tokens per day through OpenRouter. Token volume is the most honest public measure of agent usage: it counts work people actually ran, not stars or hype. One caveat: agents that route through their vendor's own API, like Claude Code and Codex, are undercounted here.

The leaderboard
Captured June 10, 2026
  • 1Hermes Agent | OpenRouter logo
    Hermes Agent | OpenRouter[Skip to content](https://openrouter.ai/apps/hermes-agent#skip) [OpenRouter](https://openrouter.ai/) / No models found [Models](https://openrouter.ai/models)[Fusion](https://openrouter.ai/fusion)[Chat](https://openrouter.ai/chat)[Rankings](https://openrouter.ai/rankings)[Apps](https://openrouter.ai/apps)[Enterprise](https://openrouter.ai/enterprise)[Pricing](https://openrouter.ai/pricing)[Docs](https://openrouter.ai/docs/quickstart) 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. [OpenRouter](https://openrouter.ai/) © 2026 OpenRouter, Inc
    Tokens/day814B
    Growth—
  • 3OpenClaw logo
    OpenClawOpenClaw is an open-source AI agent that connects to your messaging apps and takes real actions on your behalf.
    Tokens/day186B
    Growth—
  • 4Claude Code logo
    Claude CodeClaude Code is Anthropic's agentic coding tool that reads your entire codebase, plans and executes changes across files.
    Tokens/day115B
    Growth—
  • 15Codex logo
    CodexA coding agent that helps you build and ship with AI.
    Tokens/day15.5B
    Growth—

Source: OpenRouter (openrouter.ai/apps), as of 2026-06-10.

How to read this board

OpenRouter is an API gateway that serves hundreds of models behind one key, and it publishes which apps route the most tokens through it. Because tokens map directly to work performed, the board doubles as the best public census of agent usage. We capture it daily and keep the history.

Two structural caveats. First, vendor agents like Claude Code and Codex talk to their own APIs by default, so only a slice of their traffic shows up here; their true usage is far higher than their rank suggests. Second, the board moves fast: stealth launches and viral agents can climb from nothing to the top ten in weeks, which is exactly why a daily capture is worth keeping.

Frequently asked questions
What is this leaderboard based on?

It ranks apps and agents by the number of tokens they route through OpenRouter, a popular API gateway for AI models. Token volume measures real usage: every row reflects work people actually ran through that agent, not stars, downloads, or survey answers. We capture the public OpenRouter apps board daily and publish it here in a crawlable form.

Why do Claude Code and Codex rank lower than expected?

Claude Code routes most usage through Anthropic's API and Codex through OpenAI's own infrastructure, so OpenRouter only sees the slice of their traffic that users explicitly route through it. Their position here understates their real-world usage. Agents that are OpenRouter-native, like Hermes and OpenClaw, are measured much more completely.

What does tokens per day mean?

Tokens are the units AI models read and write; an agent that processes 800B tokens per day is doing roughly that much reading and writing across all its users in a single day. It is the closest public proxy for how much real work an agent performs.

How often is the leaderboard updated?

Daily. We capture the OpenRouter apps board once a day and keep the history, so ranks and token counts on this page reflect the latest captured day.

Share:
Details: