A quick visual tour of what the tool shows you.
One command reads the session logs your coding agent already saved and tells you
where it starts getting worse. Everything runs on your machine — no API keys, no uploads.
All screenshots below are from a synthetic demo dataset, so no real project names appear. Your own report uses your real sessions.
contextrot
The headline is a plain verdict — rot, edge rot, clean, or not enough data — followed by the exact context-fill % where you start failing, a failure-rate curve with confidence intervals, where your context is being spent, and concrete fixes. Use more than one model and it compares them head-to-head automatically (here Opus rots 3.8× while Sonnet stays flat on the same work).
contextrot --html report.html
One self-contained local file (still zero network) with a built-in 1200×630 share card you can save as a PNG and post.
contextrot projects
An independent curve and verdict per project, ranked worst-first — so the specific repo whose
CLAUDE.md or MCP setup is dragging you down stops hiding inside your all-projects average.
contextrot agents
Claude Code vs Codex CLI vs Gemini CLI vs Cline — each with its own curve and verdict on a shared scale, measured on your workload rather than a benchmark's.
contextrot trends
Week-over-week failure rate and startup-overhead bloat, with an honest verdict on whether the
change cleared statistical noise. This is the before/after check for contextrot fix.
contextrot fix
Turns the report's findings into concrete actions, and points out MCP servers you configured
but never actually use. Dry-run by default — it writes nothing unless you add --apply
(which backs up first and is reversible).
contextrot sessions
A plain list of everything that was read, with each session's peak context fill.
The report tells you where you degrade after the fact. These put it in front of you while
you're working — and they're the reason a Claude Code user gets the most out of contextrot.
All three are dry-run by default, write only with --apply, back up your settings first, and
undo cleanly with contextrot uninstall.
contextrot install statusline --apply
Your current context fill, colored against your own measured curve — not a generic "yellow at 70%." It knows where you start failing, and recalibrates on every run.
contextrot install hook --apply
One nudge, the instant a session crosses your measured failure threshold — then silence until the next crossing. No knee in your data? It says nothing at all.
claude mcp add contextrot -- contextrot mcp
Runs contextrot as an MCP server so Claude Code itself can pull your rot report during a session and decide to compact, warn you, or switch models. Still zero network — a local pipe, not a socket.
uvx contextrot
# or
pip install contextrot
contextrotNo config, no API keys, no uploads. contextrot makes zero network calls — local files in, terminal or local HTML out.