Show off your agent setup
We've scored 16.2k public Claude Code, Cursor and Codex setups on context tax and guardrails. Find yours, see how it compares, and add the badge to your README.
What the score means
- Context tax: an estimate of how many tokens your committed setup loads into the agent every turn (instruction files, MCP tool schemas, skill and subagent descriptions). The badge shows it with a tier: Featherweight (under 2k), Moderate (2k to 6k), Heavy (6k to 15k), Context hog (15k+). The median across all setups is ~2.3k. For Claude Code setups with MCP servers the number is an upper bound and the badge shows it with ≤ (e.g. "≤4.1k tok"): Claude Code loads MCP tools on demand, so the real cost is usually lower.
- Guardrails x/5: one point each for blocking destructive commands, protecting secrets, a pre-tool screening hook that actually checks something, no YOLO mode, and a sandbox or ask-first rules. Full method.
Scores come from the files committed to the public repo. They're estimates, not measured usage. Re-check after any change; the badge updates by itself after the next daily crawl.
Add the badge
Every indexed rig page has an "Add the badge" box with a one-click copy. The Markdown looks like this:
[](https://agentrigs.dev/r/OWNER/REPO/?ref=badge)
Optional: add ?style=flat-square to the image URL for square corners.
Remove your repo
Don't want to be listed? Use the opt-out form, or add an empty .agentrigs-ignore file to the repo root. Removals apply at the next build (usually within 24h).
What we count
To see whether badges are useful, we count badge image loads per repo per day, copy-button clicks, lookups on this page and rig page visits from a badge link. These are plain counters: no IPs, cookies or personal data are stored. The totals are public at /api/badge-stats.