Rare find
Contextrot. See exactly when your AI coding agent starts making mistakes as it runs out of memory.
pypi.org/project/contextrot
Vaya's read on this project
Problem, audience, market, and the verdict — sign in to see it.
Updates
July 2026
- Release —v1.3.0 — verdicts that work on a 1M-token window
- fix: choose fill zones from the data so 1M-window users get a verdict
- Release —v1.2.0 — correct 1M context windows + live status anywhere
- fix: correct 1M context windows; add universal status + doctor
- Release —v1.1.0 — a report a normal person can read
- feat(report): make the report readable by a non-expert + polish visuals
- style: wrap two long lines in opencode JSON adapter test (ruff E501)
- Release —v1.0.2 — actionable 'not enough data'
- feat(cli): suggest a wider --days when a run is 'not enough data'
- Release —v1.0.1 — OpenCode file-storage fix
- fix(opencode): read current file-based storage, not just legacy openc…
- docs: add visual SHOWCASE gallery, refresh media, add engagement temp…
- docs: explain the full feature set in plain words
- Release —1.0.0 — contextrot is stable
- Release 1.0.0
- Release —0.12.1 — fix MCP rate typing
- Fix nullable-rate rounding in MCP server (0.12.1)
- Update roadmap for 0.8.0-0.12.0 wave
- Release —0.12.0 — local SVG verdict badge
- Add local SVG verdict badge (0.12.0)
- Release —0.11.0 — trends: is your context hygiene actually improving?
- Add week-over-week trends (0.11.0)
- Release —0.10.0 — MCP server: your agent can now ask how rotted you are
- Add MCP stdio server (0.10.0)
- Add knee-crossing PostToolUse hook for Claude Code (0.9.0)
- Add live Claude Code statusline calibrated to your own curve (0.8.0)
- Add per-agent comparison (0.7.0)
- Add Cline / Roo Code / Kilo Code adapter (0.6.3)
- Add Gemini CLI + Qwen Code adapter (0.6.2)
- Add Codex CLI adapter (0.6.1)