Agent Usage

Rankings

Which models and coding tools builders pick, and which they actually run. Votes come from people. Usage comes from reported tokens. Neither one is a benchmark.

Most reported tokensClaude Code97.2% of tokensTop open sourceCodex CLI2.0% of tokens

Best coding tool

The agent or CLI builders reach for first. One vote per account per month.
Rank byAdd a tool
No votes this month. Until there are, tools are ordered by how much reported usage runs through them.Cast the first vote
#ToolVotesUsage shareMost run modelBuilders
1
Claude CodeAnthropicCLI
97.2%
Claude Opus 5.55
2
Codex CLIOpenAIOpen sourceCLI
2.0%
GPT-6.1 Sol1
3
Gemini CLIGoogleOpen sourceCLI
No reported usage yet——
4
Muse CodeMuseCLI
No reported usage yet——
5
OpenCodeSSTOpen sourceCLI
No reported usage yet——
6
PiPiOpen sourceCLI
No reported usage yet——
Usage share is each tool's part of reported tokens in the period, from the open-source companion. Cursor counts only accounts that verified with Cursor. OpenRouter runs inside other tools, so its tokens are counted under the tool that sent them.Vote in every category
Missing a tool?Add it and it's on the ballot right away, with your name on it. Usage share appears once the companion can read its logs.
Add a tool