Disclosure: This is written by Lorikeet Security. The Cait column sticks to what Prescient Security has published publicly - their May 2026 launch release, their Cait product page, and the July 2026 Help Net Security piece on the ASM expansion - rather than what we'd like it to say. Anything Prescient hasn't published is marked not published rather than guessed at. Re-check what's actually shipped before using this in a bake-off; Prescient's roadmap items were dated summer 2026.
The One-Line Difference
Cait is a continuous, always-on layer that keeps re-testing the same assets so coverage never goes stale between annual pentests.
Lory is a full engagement that runs on demand or on a repeating schedule, where a named human pentester signs every finding before it reaches you.
Both now give you recurring, unattended coverage. The difference that's left is what happens to a finding: Cait's model is that AI provides coverage while human testers work the hardest problems, and Lory's is that no finding reaches you at all until a person has read the evidence and signed it.
At a Glance
| Dimension | Cait by Prescient | Lory by Lorikeet |
|---|---|---|
| Model | Continuous, recurring AI-assisted service | On-demand or repeating engagements, scoped per run |
| Platform | Cacilian PTaaS platform | Portal, terminal, or MCP |
| Time to start | Recurring schedule, always on | Immediately on request, or automatically when a scheduled run comes due |
| Recurring coverage | Always-on layer, re-tests the same assets after changes | Repeating engagements on your cadence, queued and started automatically, each one signed |
| Out-of-scope handling | Not published | Held for human scope review, never run silently |
| Bring your own agent | Not published | MCP server, works with Claude Code, Cursor, or any MCP-aware agent |
| Quote | Fixed-price subscription or one-time engagement | Itemised scoped quote within 24 hours |
| Pricing model | Subscription or per-engagement fixed bid | Prepaid credits, no seats, no subscription |
| Human sign-off | Human testers focus on the hardest problems | A named pentester countersigns every finding, no exceptions |
| Evidence standard | Exploit-validated, audit-ready proof | Reproduced with request/response, screenshots, and the chain it unlocked |
| Coverage record | Not published | Vectors planned, run, and never reached, with the reason attached |
| Attack surface discovery | ASM, announced summer 2026 | Not offered, you declare the scope |
| Asset types | Web apps at GA, expansion beyond web announced for summer 2026 | Web app, API, mobile, network (internal and external), cloud, source code review |
| Physical | Not offered | Human-led, never run by Lory |
| Retesting | Continuous re-testing after changes is the core of the model | On demand against your existing credit balance, closed by a human |
| Machine-readable output | Not published | SARIF 2.1.0, GitHub code scanning, MCP |
| Framework mapping | SOC 2 and ISO called out | CWE plus control mapping across 7 frameworks |
| Methodology structure | Explores and understands the app before attacking | 104 written playbooks across 12 attack categories, one pass per vector |
Coverage, Side by Side
This is the widest gap between the two products, and the easiest one to demonstrate.
| Asset Type | Cait | Lory |
|---|---|---|
| Web app | Yes, the launch focus | Yes |
| API | Not called out separately | Yes, crawl, auth testing, injection, access control |
| Mobile | Not published | Yes, iOS and Android plus their backends: storage, IPC, deep links, pinning |
| Network, external | Not published | Yes, services, versions, exposure |
| Network, internal | Not published | Yes, internal ranges via the mesh connector |
| Cloud | Not published | Yes, AWS, Azure, GCP, Kubernetes, containers, serverless: IAM, metadata paths, privilege escalation |
| Source code review | Not published | Yes, secret hunting and SAST-style sink tracing, plus supply chain |
| Physical | Not offered | Human-led, never run by Lory |
Cait went GA web-app-focused, and Prescient announced in July 2026 that it would extend to more asset types over the summer. Check what has actually shipped before you put "web app only" in front of a prospect, because that claim has a short shelf life and getting it wrong in a bake-off is expensive. The safer framing is what you cover, not what they don't.
Worth Saying Plainly
Because a comparison that pretends otherwise isn't useful to anybody.
- Continuity is the whole product. Cait is architected as an always-on layer rather than a series of runs, and it's the story Prescient leads with. Lory now runs repeating engagements automatically, so the practical gap here is small, but Cait owns the word "continuous" in the market and that carries weight with auditors who read press releases.
- ASM closes the loop. Discovering what's exposed, then testing it, then tracking remediation, all in one platform, is a real workflow advantage. Lory starts from a scope you declare, which means anything you forgot to declare doesn't get tested.
- Platform maturity. Cacilian exists as a unified PTaaS platform with structured intake and real-time visibility, backed by an established audit and cybersecurity firm with named leadership.
- Predictable budgeting. A fixed-price subscription is easier to get through procurement than a credit balance in some orgs, even though credits are cheaper for bursty usage.
Where Lory Differentiates
- Nothing ships unsigned. Prescient's framing is that AI provides coverage while human testers focus on the hardest problems. Lory's is stricter: a human reads the evidence, rates the risk, and countersigns every finding, and only a human can close one. That's a different accountability contract, not a different amount of automation.
- A coverage record you can read. Every run tells you which vectors were planned, which ran, and which were never reached and why. Neither scanners nor most pentest firms will tell you what they missed. We haven't seen Prescient publish an equivalent.
- Scope as an enforced gate. Targets and rules of engagement are signed up front and every tool call is checked against that allowlist before it fires. This matters for anyone who has to defend an autonomous tool's behaviour to a regulator, a customer, or their own legal team.
- One vendor covers the whole estate. Web, API, mobile, internal and external network, cloud, and source code review all run through the same engagement, the same credit balance, and the same signature. A buyer standardising on Cait for web apps still needs somebody else for their internal ranges, their mobile app, and their cloud IAM.
- Findings leave the PDF. SARIF 2.1.0 and MCP mean findings land in GitHub code scanning and in your agents, not just in a report nobody opens twice.
- Chains, not isolated findings. Each finding shows the path from a small misconfiguration to the thing you actually care about, rather than a list of independent items.
- Recurring coverage without a subscription. Set a cadence - weekly, per release, quarterly - and Lory queues each run and starts it automatically when it comes due. Pause or cancel a schedule at any point. Runs draw from your credit balance, so you get standing coverage without a seat count or an annual floor.
- Hook up your own agent. An MCP server puts your workspace in reach of Claude Code, Cursor, or anything else MCP-aware, so engagements and findings are available to the tools your engineers already run.
Honest Weak Spots for Lory
Cait says "continuous," Lory says "recurring" - functionally close now, but a buyer working from a compliance checklist may pattern-match on the word. There's no ASM: scope-as-a-gate is a security property, but unknown assets stay untested, and a buyer who doesn't know their own surface gets more from Cait's discovery loop. Human countersign is a throughput ceiling by design - it's the feature and the constraint, and at high volume it's the bottleneck a prospect will ask about. Prescient is also a known audit and security firm, and that carries weight in procurement independent of product quality.
Picking Between Them
Cait may fit if...
- Your scope really is just web applications
- You don't have a clear picture of your external attack surface and want discovery bundled in
- Predictable subscription billing suits your finance team
- An established audit-firm name carries weight in your procurement
Lory may fit if...
- You have more than web apps to test - API, mobile, network, cloud, or source code
- You need findings a named person stands behind
- You want proof your engineers can reproduce in minutes, plus the chain it unlocked
- You want a written record of what wasn't tested
- You'd rather run a standing cadence off a credit balance than carry a subscription
They aren't mutually exclusive. A continuous layer and a signed engagement solve different problems, and some orgs will end up running both.
See a Signed Engagement in Action
Book a scoping call and we'll walk you through exactly what a Lory engagement looks like end to end - the playbooks, the evidence standard, and how a human pentester signs off before anything reaches your report.
Book a Consultation
-- views
Link copied!
