Skip to main content

orgx trail · open source · runs on your machine

See the mistakes your coding agents keep repeating

Trail reads the Claude Code, Codex, Cursor, Copilot, Gemini CLI, OpenCode and Factory Droid history already on your machine. It groups the failures your agents hit again and again, shows how many sessions hit each one, and drafts the rule that stops it. You choose what goes into your agent instructions.

npx -y @useorgx/trail

No history yet? npx -y @useorgx/trail --demo runs on made-up sessions. Or ask your agent: “Run npx -y @useorgx/trail scan, then trail walls, on this machine. Show me the repeated failures and the suggested rules. Don’t upload anything or change any config until I say so.”

No account · Node 20+ · The browser view runs on localhost · Only counts leave your machine, and only if you share them

Rules other people adopted

Each fix below is one rule for AGENTS.md or CLAUDE.md you can paste today. Before-and-after numbers appear once 5 people have measured it on their own machines; until then the row shows how many have.

How agents are allowed to work

The machine they run on

How it’s measured: trail compares sessions in the same repository and the same permission mode before and after the rule was added. When the permission mode changed, the comparison is left out instead of claimed. Only counts leave a machine; never code, prompts or file names.

Trail fixes your machine. OrgX fixes your team.

A rule in your CLAUDE.md fixes one repo on one laptop. Connect OrgX and a rule you confirm reaches every teammate and every connected tool, and every result is checked against it. Start free → Built on the work of others →