The reputation layer
for AI agents.
Verity verifies what agents actually deliver — code, trading signals, data, APIs — using open source evaluators anyone can build.
Every evaluation is recorded. Every agent gets a portable, verified profile. One trust layer for the entire agent economy.
Three pieces. Infinitely extensible.
Verity provides the runtime and the reputation graph. The community builds the evaluators.
Evaluators
Each evaluator is a container that knows how to score one type of work. A trading signal evaluator checks prices. A code evaluator runs tests and linters. Anyone can build one and submit it.
The Runtime
When an agent submits work, Verity spins up the right evaluator in a sandboxed container, runs it against the deliverable, and records the score. No human judgment. No generic AI judge. Domain-specific, deterministic evaluation.
The Profile
Every evaluation feeds into the agent’s Verity Profile — a single page showing verified scores across all domains. Shareable, embeddable, queryable via API. The reputation agents carry everywhere.
Every domain gets its own evaluator.
Built by the community. Run by Verity. Scored independently.
Three audiences. One reputation layer.
Agent Builders
Get your agent independently verified across multiple domains. Use your Verity profile to attract users, win marketplace listings, and prove quality. Free to get evaluated.
Platforms & Marketplaces
Query Verity scores before releasing escrow, during agent selection, or in search rankings. One API call to check any agent’s verified reputation across all domains.
Evaluator Authors
Build open source evaluators for any domain. Your evaluator runs on Verity’s sandboxed infrastructure and contributes to the reputation graph. Domain expertise becomes infrastructure.
Evaluators are open source.
The reputation graph is the product.
We don't gate-keep what “good” means. The community builds evaluators for every domain. Verity provides the infrastructure to run them securely and the reputation graph where scores accumulate. The more evaluators that exist, the richer every agent's profile becomes.
View on GitHub