FoundryNet / Solutions / Agent trust verification

AGENT GOVERNANCE · TRUST & REPUTATION

Your Agent Has an Identity. Does It Have a Reputation?

60% scaling agents·4% governing them·Scored on work, not claims

Identity tells you which agent it is. Reputation tells you whether it should be allowed to do this. Assay scores agents on completed, verified work.

Get Forge on AWS Marketplace →See pricing →

Assay is the governance layer beside Forge. Forge normalizes the machine; Assay scores the agent acting on it.


The problem

Identity is solved. Competence is not.

Around 60% of enterprises are scaling AI agents into real workflows. Roughly 4% have a governance model for them. The gap is not authentication — issuing an agent a credential is a solved problem. The gap is that a valid credential says nothing about whether the agent is any good at the task it just requested.

In an industrial setting that distinction has teeth. An agent authorised to adjust a process setpoint is authorised because someone once decided it should be. Nothing in that decision reflects its track record since, and nothing stops a capable-sounding agent from claiming a capability it has never demonstrated.

60%Enterprises scaling agents
4%With agent governance
0Self-claimed capabilities scored

How Assay scores

Reputation earned from work, not asserted in a profile

Scored work history

Trust is derived from completed, verified work records — what the agent actually did and whether it held up — not from a self-declared capability list.

Verification ceiling

An agent's score cannot exceed what its verified work supports. Volume alone does not buy trust; unverified volume buys none of it.

Velocity discount

Reputation is wall-clock work. Bursts of activity are discounted against the median gap between records, so trust cannot be manufactured in an afternoon.

Cross-org work cells

Agents from different organizations can coordinate inside a shared cell, each carrying its own scored history rather than inheriting the host's trust.


Using it

Trust-gated coordination and Agent Passports

A trust gate is a threshold on an action: this operation requires a score above x, from an agent with verified history in this domain. An agent below the line is refused before it acts, not audited afterwards.

An Agent Passport is a signed, time-boxed summary of an agent's identity and standing that another organization can verify without querying your systems or trusting your word for it. Platform templates are available for common agent runtimes.

Agent identityTrust scoreWork recordsCross-org cellsTrust gatesAgent Passport
Get Assay on AWS Marketplace →Open the Assay dashboard →

Related

Keep reading

Industrial MCP server

The 32 tools those agents call once they are trusted.

The missing data layer

A well-governed agent still needs input it can reason about.

Equipment finance intelligence

Verifiable records matter most where the counterparty has an incentive.


From the newsletter

The Machine Agent

Longer-form writing on industrial AI infrastructure by Foundry Labs.

The Agent Passport Problem

The trust crisis hiding in plain sight

Read on Substack →
The Missing Layer

Why credit scores for agents is the wrong framing, and what actually matters

Read on Substack →
Work Cells

The unnamed primitive powering the machine labor economy

Read on Substack →