C
Canonopy Labs
Open beta — sign up, get your key, start in minutes

Agents witha warranty.

Most agents never leave the pilot, because nobody can say what they'll never do. Everyone else logs what your agent did — we certify what it won't. Not a wrapper with a stern system prompt: a research technique we developed that turns your written policy into a decision layer whose behavior can be exhaustively verified — your agent is tested against thousands of adversarial cases built from it, and you can re-run that certification yourself, seed for seed. In production, a forbidden action can't get through the guard — and every ruling ships with a receipt citing the exact rule that decided it. The technique is ours; every result is yours to re-run. One team's violation rate: 68% before, zero after.

Certification batteryre-runnable
12,000 / 12,000
adversarial cases · 0 violations
PASS — GUARANTEED
Decision receipt#a41f · 84 ms
agent proposed: process_refund($482)
order: shipped · identity: unverified
DENIED
"Refunds require identity verification and may only be issued while an order is pending." — your policy, §4.2
returned: verify_identity ALLOWED
Audit — last 30 days
214
violations found in 3,100 logged decisions, each with the policy line it broke
▸ The cards above are the three artifacts you get — battery report, decision receipt, audit summary — in the engine's own output format. Run the battery yourself on day one.
Why agents are stuck in review
88%
of enterprise agent pilots never reach production — behavior nobody can bound doesn't clear legal and security review
<70%
what frontier agents score on the standard public policy-following benchmark — prompting isn't a control
Aug 2026
EU AI Act logging and human-oversight obligations for high-risk AI systems became enforceable — "we have logs" is no longer an answer
What changes with a warranty
0
forbidden actions can execute through the guard — by design, not statistics
100%
agreement with the industry agent benchmark's executable environment
8
errors our engine found in that benchmark's own answer key
68→0%
a test agent's violation rate, before and after one integration step

Grade. Guard. Generate.

Three steps, in the order your team will actually adopt them — each works with the agent you already have.

1

Audit what it already did

Point your coding agent at last month's logs — it extracts the facts of each decision, our engine grades them. Every decision comes back marked compliant or violation, with the exact policy line and the correct action. Your first compliance report, in minutes.

audit(logs) → 214 violations · citations · fixes
2

Make it safe without changing it

One call before each action. Legal ones pass. Illegal ones can't execute — your agent gets the citation and the right alternative to relay. Brain untouched; hands only bend in legal directions.

guard(action, case) → allow | deny + citation
3

Deploy the certified layer

When you're done re-teaching rules your LLM keeps forgetting, the decision layer takes the wheel — certified by a battery of thousands of adversarial cases you can re-run any time.

create_agent(policy) → certified · receipts on

It lives where you already work

No SDK, no migration. Developers use it from their coding agent; production calls a plain HTTPS API; compliance gets a read-only receipts dashboard.

Claude Code / Cursor — dev time

claude mcp add --transport http warranty \
  https://api.canonopylabs.com/mcp \
  --header "X-API-Key: $KEY"
Then just say: "audit our agent's logs against refund-policy.md" — your own assistant does the rest.

Your agent's runtime — production

POST /v1/guard
{ "action": "process_refund", "case": {…} }
→ { "allow": false,
    "denied_because": ["your §4.2"],
    "suggested_action": "verify_identity" }
Deterministic · median well under 100 ms · every call receipted.

Built on research, not vibes

This isn't a wrapper around a model with a stern system prompt. It comes out of a research program on making agent behavior provable — a technique we developed that turns a written policy into a decision layer whose behavior can be exhaustively verified, and that makes the agent itself trainable against that layer. The technique is ours; every result it produces is yours to re-run.

Measured, not claimed

One method, five unrelated domains — commerce support, airline travel, telecom support, robot navigation, and reply-quality vetting — certified with the same machinery. Checked against two generations of the industry's standard agent benchmark by replaying decisions through their own executable environments: 100% agreement, and 8 errors surfaced in one hand-written answer key along the way (graded write-up available on request). And it's predictable: accuracy on a new domain follows a curve we've measured across unrelated domains, so we can tell you the onboarding cost for a target accuracy before we start — not after.

Verifiable by skeptics

Every number on this page regenerates from a seed: run the battery yourself, hand the seed to your auditor, get the identical report. The research stays behind the API — what crosses the boundary is rulings, receipts, citations, and labeled training examples. Your LLM can't leak what it never receives; neither can ours.

"But our rules are subjective"

Most aren't — and the ones that are split cleanly into three tiers. We tested this: a reply-quality policy full of subjective words — empathy, tone, dismissiveness — became a fully certified domain (0 violations across 2,000 cases), and a 21KB model learned the entire rule set to 100% held-out accuracy from exported examples alone.

1

Articulable → certified

"Never blame the customer." "No unapproved promises." "Every reply needs a next step." These sound like judgment calls; written down, they're checkable rules. Most of what teams call subjective lives here — during onboarding it becomes part of your certified rule set, and training data for your own model.

2

Ambiguous → adjudicated

"Show empathy when the customer is upset." Does acknowledgment count, or does it take an apology? Both readings are defensible — so the sign-off surfaces the split, the policy owner picks one, and from that moment it's tier 1: a certified rule with a citation. Ambiguity gets converted, not averaged.

3

Ineffable → guarded

How warm the wording feels, whether a joke lands — no rule decides that, so we don't pretend to. Your model keeps its voice; the guard underneath makes sure nothing it says can turn into a forbidden action or an unapproved commitment. Free on the surface, bounded at the decision.

Straight answers

The questions a serious buyer should ask — answered the way we'd answer an auditor.

What exactly is guaranteed?

The decision. Given the facts of a case, the ruling is deterministic: the same facts always produce the same allow/deny, the receipt names the policy line that decided it, and the certification battery re-runs identically for you or an auditor, seed for seed. Upstream of the decision, your agent extracts those facts from the conversation — that step is language, not law, so the facts the engine ruled on are printed on the receipt. If extraction got something wrong, the receipt shows it. That's the honest boundary: rulings are guaranteed; everything that fed them is auditable.

How do I know your rules actually match my policy?

Because you approve decisions, not prose. At onboarding you sign off on dozens of concrete decided scenarios — case, ruling, citation — which is where ambiguities in a written policy surface before production, not after. (They exist in almost every policy: a public benchmark's own policy text yielded one.) You can add your own gold cases to the battery at any time. And the approach has been checked against an independent referee: on a public agent benchmark, our rulings were replayed through the benchmark's executable environment — 100% agreement — and the 8 disagreements with its hand-written answer key turned out to be errors in the answer key. Write-up available on request.

I'm in compliance or risk, not engineering — what do I get?

Three artifacts, none of which require trusting us. A certification you can re-run yourself — same seed, identical report, any time, including in front of an auditor. A receipt for every production decision, showing the facts the engine saw, the ruling, and the policy line that decided it — retained per your tier, so "what did it see at decision time?" has an answer months later. And an audit trail where denials are first-class records, not absences — you can show what the agent was stopped from doing, with the rule that stopped it. If EU AI Act obligations are on your desk: the logging and human-oversight evidence it asks for is what this system produces as a side effect of running.

Why not Cedar, OPA, or another policy engine?

Use them — for infrastructure authorization; they're good at it. Cedar and OPA are rule evaluators: you hand-write rules in their language, they answer allow/deny, and that's where they stop. This layer starts from your policy in plain English, handles the stateful business rules authorization languages express awkwardly — order lifecycles, booking windows, refund routes — returns receipts that cite your own wording, ships an adversarial certification battery you can re-run, and exports labeled training examples that make your own model better between audits. One is a permissions component; this is the compliance loop around your agent.

Pricing that starts where you are

Free to find your violations. Thirty dollars to guard your first agent. The serious tiers arrive when the compliance questions do.

Free
$0
  • 1 policy
  • Audit 1,000 decisions / mo
  • 100 guard calls / mo
  • Battery on demand
Start free
Startup
$30/mo
  • 2 policies · unlimited audits
  • 10,000 guard calls / mo
  • 7-day receipts
  • Training-example exports
Start — $30/mo
Team
$299/mo
  • 10 policies · 5 seats
  • 50,000 guard calls / mo
  • CI battery gate on deploys
  • 6-month receipts retention
Become a founding partner
Enterprise
Custom
  • Hosted certified agent
  • Unlimited retention + audit packs
  • SSO · SLA
  • Templates built for your domain
Talk to us

Create your access key

Built for teams with an agent stuck in review — or one they'd rather never review again. Free tier, no card. Your key works on both doors: the MCP endpoint your coding agent connects to, and the REST API your runtime calls.