Convex Advisor

Point convex-advisor at a live Convex deployment and it reads the last 72 hours of insight events from the official Convex MCP, things like hard read-limit hits, approaching-threshold warnings, and optimistic-concurrency retries on hot documents. Each event gets traced back to the function's source, so you get a finding like "messages:list read 4.2MB from messages 31 times yesterday" with the file and line responsible and a concrete fix. It only reads, never mutates prod, and it deliberately refuses to guess when the deployment has no traffic to learn from. The honest catch is that it needs cloud dev or prod access while logged in as a user, plus roughly three days of real traffic, so it's useless on fresh projects or preview deployments. Where it earns its keep is catching the expensive unindexed scan you didn't know was running.

skills.sh installs: 41.1K
GitHub stars: 69
last updated: Oct 26, 2026

View on GitHub

Install

$ npx -y skills add convex/agent-skills --skill convex-advisor --agent claude-code

Installs into .claude/skills of the current project.

Files

Live-deployment advisor

Static review guesses; the deployment KNOWS. The official Convex MCP ships an insights tool with typed 72h health events per function — documentsReadLimit / bytesReadLimit (hard limit hits), documentsReadThreshold / bytesReadThreshold (approaching), occFailedPermanently / occRetried (write contention) — each carrying evidence (table_name, bytes_read, documents_read, occ document id + retry count). The advisor turns each event into a root-caused finding by reading the flagged function's actual code, and emits findings on the findings bus (specs/finding.schema.json) so fixers can be dispatched and launch-readiness can score.