Skip to main content
AV

Independent verification for AI agents and their value

av9n scores AI agents with a tamper-evident record, so you can show not just that work ran, but whether it was worth the cost. It is aimed at builders and teams that need auditable proof of value rather than basic observability.

Agenticness = how independently a tool can take action, scored across 9 dimensions. Scored independently by David Kooi, Skylark Creations — see full rubric →

See pricing ↓

Enterprise
API
B2B
For Developers
For Teams
Visit av9n

Ask about av9n

Get answers based on av9n's actual documentation

Try asking:

About

What It Is

av9n is a verification and measurement layer for AI agents, not an agent itself. It is built for developers and enterprise teams that need to prove whether an agent produced net value after accounting for costs like tools, infrastructure, tokens, and human review time.

What to Know

av9n is useful when you need an auditable, externally checkable record of what an agent did and what that work was worth. The strongest part of the product is its emphasis on tamper-evident evidence, counterfactual baselines, and cost-aware scoring....

Key Features
Wraps agent tool calls with a Python decorator
Creates tamper-evident sealed records for each agent action
Generates a health score from value, assurance, and cost
Compares outcomes against a baseline or live holdout
Splits credit across multi-agent chains to avoid double-counting
Use Cases
Audit an AI agent before shipping it to customers
Show a buyer a signed evidence bundle proving agent value
Track whether an internal workflow agent is still producing net benefit
Agenticness: Reactive Tool

Responds to prompts but takes no autonomous action.

High evidence
Last evaluated: Aug 3, 2026

Dimension Breakdown

Action Capability
Autonomy
Planning
Adaptation
State & Memory
Reliability
Interoperability
Safety
Operator Sovereignty

Categories

Pricing
  • Free: Developer tier at $0/month with 250 verified scores per month, full SDK and sealed ledger access, a health score plus four-question breakdown, directional grade only, and community support.
Details
AddedAugust 3, 2026
RefreshedAugust 3, 2026
Agenticness
Quick Facts
DeploymentCloud-hosted
AutonomyCopilot (human-in-loop)
Model supportSingle model
Open sourceNo
Team supportEnterprise
Pricing modelFreemium
Interfaceapi, gui, web
Sources
Last updated August 3, 2026
Stay in the loop

Get the weekly agentic AI briefing

New tools, top picks, and trends — delivered every Thursday.

I use AI for: