Blume
A desktop sidecar application that monitors, manages, and optimizes multiple AI coding agents (Cursor, Claude Code, Codex, omp, Pi) by tracking rules, skills, and configuration patterns. Detects repeated corrections and suggests systematic improvements wh
🔗 Visit Blume
Description
If you use more than one AI coding assistant — say, Cursor for one project and Claude Code for another — you've probably noticed you end up repeating the same corrections to each of them over and over, because they don't share what they've learned. Blume watches for that pattern and turns your repeated corrections into a reusable rule or skill, so you stop teaching the same lesson twice.
Blume is a desktop sidecar application that monitors and manages multiple AI coding agents (Cursor, Claude Code, Codex, omp, Pi) from one place. It tracks each agent's configuration files (rules, skills, hooks), shows token consumption and plan usage per agent, previews code diffs before applying suggestions, and detects patterns in repeated corrections to suggest new rules or skills — all with conversation history stored locally for privacy.
💬 Our review
The short version: Blume is solving a genuinely annoying problem — repeating the same correction to five different coding agents — but it only pays off if you're actually running multiple agents side by side.
There's no direct competitor doing exactly this: Cursor's rules system and Claude Code's skills/hooks each manage configuration within their own tool, but nothing else unifies monitoring and pattern detection across multiple agent platforms simultaneously, which is Blume's real differentiator. The catch is that pricing isn't disclosed on the website, and the value proposition collapses if you only use one coding agent — you're paying (in setup time, if nothing else) for cross-agent oversight you don't need. If you genuinely juggle Cursor, Claude Code, and others day to day, Blume's pattern detection can save real time by systematizing your corrections instead of repeating them; single-agent users should just use that agent's native rules/skills system instead.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Tarification non communiquée sur le site ; accès gratuit disponible via le lien Get Blume, modèle freemium probable avec fonctionnalités d'équipe payantes à venir.
Pros
Supervision unifiée entre plusieurs plateformes d'agents de code, élimine le changement de contexte
Stockage local axé sur la confidentialité, évite la fuite de données vers des services externes
Détection intelligente de motifs pour automatiser la découverte d'optimisations
Suivi en temps réel des tokens et de l'usage, évite les dépassements de plan imprévus
Cons
Une application supplémentaire à apprendre et intégrer au workflow existant
Modèle tarifaire flou, surprises possibles à la mise à niveau
Support de plateformes limité par rapport aux agents autonomes
Nécessite une configuration multi-agents pour offrir toute sa valeur