Keydris
Authorization layer that checks every AI agent action against a policy before it runs, with per-action revocation and a full audit trail.
🔗 Visit Keydris
Description
"Can this AI agent actually do the thing it's about to do?" is a question most agent deployments never formally ask — the agent typically just has an API key with broad access and does whatever the model decides. Keydris inserts a real permission check into that gap: before an action executes, it's evaluated against a policy, not just waved through because the agent holds valid credentials.
Keydris is authorization infrastructure for AI agents: every action is verified against operator-defined policy scope before execution, with policy versioning so rules can evolve without breaking existing integrations. It supports revoking an agent's authority mid-task without killing the whole process — useful when you need to cut off one specific capability rather than shutting the agent down entirely — and keeps decision records with full audit trails for every check performed. It works across Claude, MCP, and general APIs, with a CLI for setup. Pricing starts free for a single agent, $20/month for up to three agents (Builder), $99/month for the Pro tier, and custom Enterprise pricing above that.
💬 Our review
The short version: Keydris does one specific thing — gate agent actions behind an explicit policy check — and does it at a price point ($0-$99/mo) that individual developers and small teams can actually afford, unlike the enterprise-tier competitors emerging in this space.
It sits in the same new category as Oconee Runtime and similar AI-agent governance tools, but scoped and priced differently: Keydris is per-action authorization (can this specific action happen right now) rather than a full enterprise monitoring/SIEM platform, and it's accessible at $20/month rather than requiring a $750+/month enterprise contract. The revocation-without-termination feature is a genuinely useful detail — most simplistic "kill the agent" approaches are too blunt when you only need to cut off one capability mid-task. Being closed-source with only an organization GitHub page (not a public repo) means you can't audit the enforcement logic yourself, which matters for a security-adjacent tool — worth asking directly about this before trusting it with anything critical. Solid pick for individual developers or small teams wanting real authorization checks without enterprise-tier cost; verify the audit/compliance story yourself if regulation is a factor.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit (1 agent), 20$/mois (3 agents), 99$/mois (Pro), Enterprise sur devis
Pros
Vérification par action avant exécution, pas juste une clé API globale
Révocation d'une capacité précise sans tuer tout le processus
Piste d'audit complète des décisions
Prix accessible dès 20$/mois (vs enterprise-only chez certains concurrents)
Cons
Code fermé, logique d'enforcement non auditable publiquement
Catégorie de produit encore jeune, pas de concurrent direct établi
Moins de fonctionnalités enterprise (SIEM, SSO) que Oconee Runtime