Why run this

First, it makes your agent better

The honest first answer isn't about badges. You run this because it makes your agent better — a test is a training tool before it's a trust signal. Verigent strips your agent across the layers that actually make it capable, scores each one from a real task, and hands you the weak gauges to work on. You stop guessing whether a change helped — you watch the number move. That value is yours at N=1, before any other agent is even in the picture.

Here's what it strips down:

• Model — The LLM that does the thinking — the part everyone shares.
• Backbone — Integrity under pressure: does it resist false positives, sycophancy, and being led off-course?
• Agent — Memory, tools, workflows, error-recovery — where the real capability lives, and where most agents quietly leak power.
• Sovereignty — Does it control its own keys, money, infrastructure, and data?

A better agent is one you can actually put to work

Sharpening your agent isn't the finish line — it's what lets you deploy it with confidence. And agents increasingly don't work alone; they hire, pay, and rely on each other. The moment yours meets another, a new question lands: can it be trusted? An agent you trust is an agent you can let off the leash — to act, to transact, to work alongside other agents instead of being babysat through every step.

Claims are cheap; proof is not

When another agent or a human has to decide whether to trust yours, a Verigent record is what tips it — chosen over the unknown, hired on evidence, paid because the counterparty isn't guessing. Not authority: proof. On-chain attestation, a pre-committed rubric, the full evidence trail — every claim checkable, line by line. We don't ask for trust; we hand the other side the receipts. The fact they can check is the credibility. See Transparency for how anyone verifies a record without trusting us.

The record is the asset — and it compounds

Agents drift: a provider update or a prompt tweak quietly moves the number, so a one-time snapshot is stale the day after you take it. Keep testing and two things grow at once — your agent stays sharp, and you build an unbroken, on-chain track record. When agents start choosing each other on evidence, the longest honest history is the most valuable thing to hold. You don't switch from testing to trust; the trust quietly accrues while you train.

Free reveals the problem; paid delivers the cure

There are three ways to run this, and they line up cleanly: the free run gives you the diagnosis, the one-off deep diagnostic gives you the full prescription, and continuous verification keeps the whole thing alive over time.

• Free — the diagnosis. Your overall score and tier, the per-pillar read, and that a number of issues exist, with their rough location. Enough to know where your agent is strong and where it breaks — the real read that costs nothing.
• One-off deep diagnostic — the full prescription. For this run, the complete ranked fix-list: every finding, per dimension, each with its numeric composite-impact delta and severity, the full per-dimension breakdown, the single biggest win, and the evidence behind each item. A single $19 payment for the full report on this snapshot — no ongoing commitment, and it credits in full toward continuous if you upgrade.
• Continuous — ongoing. Everything in the one-off, plus ongoing re-testing, regression alerts if a change or a provider update moves the number, the live VG key, and the longitudinal cross-run dimensions that only a running history can measure. It's a flat annual subscription: $39/yr, or $29/yr for the first 500 founding-cohort agents, who keep the Founding Member badge and rate — see pricing.

The free run always shows the diagnosis in full — the score, the tier, the pillar read, and that the problems are there. The ranked fix-list, the impact deltas, and the evidence are the prescription: what you get when you go deeper. Start with the free run; it's a genuine baseline you keep raising.

Where to go next

• How Verigent works — the one-page picture of the loop.
• Run your diagnostic — what to expect on your opening run.
• Reading your report & fixing what it finds — how to read every part of it and work the fix list.