Agent registry

One registry for every AI agent in your organization

Register agents with framework, model, owner, and risk tier. Search by natural language. See the full dependency graph. Enforce lifecycle gates from draft to production.

Request a demo Speak with an expert
Features

Built for organizations where agents ship faster than anyone tracks them

Register in under two minutes

Add an agent with its full technical stack, owning team, and deployment URL. Agent slugs are unique per org. Every agent starts in Draft and must pass lifecycle gates before it reaches production.

See registration in action
Register agent
Name billing-reconciliation-agent
Framework LangChain
Model gpt-4o
Owner Platform Team
Tags billingfinancepii
Draft Register agent

Four-dimension risk classification

Classify across data sensitivity, decision authority, blast radius, and regulatory exposure. The system computes a composite risk tier with configurable weights. Tier drives which compliance frameworks apply.

See compliance integration
Risk classification Tier 3, High
Data sensitivity 85
Decision authority 72
Blast radius 60
Regulatory exposure 90
Composite score 76.8, EU AI Act Article 10 applies

Live dependency graph

Map how agents connect to model providers, data sources, external APIs, and other agents. Color-coded by risk tier. When a provider changes an API or deprecates a model, you see every agent affected before they break.

See the graph
billing triage notify GPT-4o Postgres Stripe
High risk Medium risk Low risk / provider

Dependency intelligence, not just a graph

The dependency graph doesn't just show connections. It monitors them. When a model provider announces a deprecation, Roval surfaces every affected agent, sorted by risk tier, with days until impact. When a framework releases a breaking change, you see the blast radius before anything breaks. When an upstream agent is deprecated, every downstream agent is flagged for review.

See lifecycle management
Dependency alerts 3 active
Claude 3.5 Sonnet deprecation, 90 days 14 production agents affected · 3 Tier 3
LangChain 0.3 → 0.4 breaking change 8 agents on affected version
data-warehouse API v2 sunset, 180 days 2 agents affected · both Tier 1
Model version tracking, framework deprecation alerts, API sunset detection, and upstream agent dependency monitoring. All from the registry, all before anything breaks.

Semantic search across your fleet

Every description is embedded as a vector. Search by what agents do, not just what they're called. Find the agent that handles PII redaction without knowing its name. Results ranked by semantic similarity.

Try a search
agents that handle PII redaction in customer data
4 results · semantic match
pii-redaction-service High
Redacts PII from inbound customer support tickets before routing
data-masking-agent Medium
Masks SSN, email, and credit card fields in analytics exports
consent-check-bot Medium
Validates consent flags before passing data to downstream agents
The problem

Agent inventories fail the moment they leave the spreadsheet

Every AI team thinks they know what agents they have. They find out they're wrong when an agent touches customer data without authorization or keeps running months after its owner left.

Without a registry
  • Agent inventory lives in spreadsheets last updated weeks ago
  • Nobody knows which agents access customer data
  • Orphaned agents run months after their creator leaves
  • No dependency mapping, one API change breaks unknown agents
  • New agents reach production with no review
With Roval
  • Every agent registered with owner, model, risk tier, and status
  • Data access classification drives governance requirements
  • Orphaned agents flagged within 24 hours of owner departure
  • Live dependency graph shows every connection
  • Lifecycle gates block uncertified agents from production
How it works

From shadow agents to governed fleet in four steps

Register

Add agents manually via the UI or auto-discover them through webhook connectors. Each agent gets a unique slug, an assigned owner, and a Draft status.

Classify

Risk tier is computed across four dimensions: data sensitivity, decision authority, blast radius, and regulatory exposure. Weights are configurable per organization.

Govern

Lifecycle gates enforce certification requirements before an agent can advance to production. High-risk agents require explicit sign-off. Uncertified agents are blocked at deploy time.

Monitor

Drift detection runs every 15 minutes. If an agent's dependencies change, its owner changes, or its certification lapses, you're notified before it becomes a compliance finding.

Platform

Registry is one module. The platform does more.

Compliance & Certification

Map agents to EU AI Act, SOC 2, HIPAA, or ISO 42001. Continuous drift detection. Export audit-ready evidence in one click.

Explore compliance

Observer & LLM Monitor

Capture every tool call and LLM request. Detect PII. Enforce policies in real time. Alert on violations before they become incidents.

Explore observer

Dashboard

Total agents, compliance posture, active violations, and LLM spend in one view. Filter by team, framework, or risk tier. Share with your board.

Explore dashboard

Stop discovering agents by accident

Join the private beta. Full registry setup takes under 10 minutes.

Request early access