Numbers, sources and how this was put together
Numbers (Q3 FY26 unless stated)
| Item | Figure | Source |
|---|---|---|
| ARR | $819.7M, −2.2% YoY | [1] |
| Q4 ARR guidance | ~$812M | [4] |
| Revenue | $212.4M, −2.5% YoY | [1] |
| Core platform | ~85% of ARR, +~1% YoY | [2] |
| Detection & response | >55% of ARR, +~5% YoY | [2] |
| Exposure management | −~6% YoY | [2] |
| Non-core | ~15% of ARR (~$123M*), −~19% YoY; end of sale for selected standalone versions | [2] [5] |
| Operating margin | 21.7% non-GAAP; ~4.2% GAAP* | [1] [2] |
| Non-GAAP gross margin | 72.2%, expected to moderate with AWS and AI compute costs | [2] |
| Cash and securities | $731.6M | [1] |
| Convertible notes | $600M due Mar 2027, to be repaid from cash | [2] |
| Customers | ~11,800 | [2] |
| Autonomous SOC | First generally available release around RSAC 2027, described as a first production step | [3] |
| Kenzo Security | Acquired 26 Mar 2026; reported customer results: 94% less investigation time, alert coverage 12% → 100% | [6] |
* My arithmetic: 15% × $820M; GAAP income from operations $9.0M ÷ revenue $212.4M. The 21.7% figure is non-GAAP, as reported; GAAP is shown alongside. Kenzo's figures are reported claims, not Rapid7-run results.
Reasoning, briefly
- Fragmented data. The homepage lists seven product lines [15]. Kenzo's approach is to integrate, normalise and unify data so agents can analyse it [6]. The same layer could be offered read-only to customers' own agents. Rapid7's open-source MCP server covers exposure data [7]; detection and response is the larger part of ARR.
- Non-core. On the call: "The issue is not quality, it is strategic fit" [5]. A possible move, moved here from the one-pager: where migration is the path, a migration offer and date for each end-of-sale customer.
- Short window. Approval controls and shadow-mode results give a cautious buyer evidence about the first release. Cost per investigation matters because gross margin is expected to moderate as AI compute grows [2].
Method
- Public information only, read 15 November 2026. No private or confidential information.
- Headline figures come from the results release and match the 10-Q [8]. Segment mix, customers and the notes come from the call transcript. Q4 ARR guidance and RSAC timing come from coverage of the call and are cited to it.
- Where no credible public figure was found, no number is used. Analyst forecasts and vendor-commissioned surveys are labelled as such.
- Three AI models reviewed an earlier draft through OpenRouter:
openai/gpt-6.1-sol-pro,anthropic/claude-fable-5.2,google/gemini-4-argon. Each got the same facts and was told to mark anything else as speculation. All three rated approval controls and an audit log the strongest idea, rated a reseller offer the weakest, and pointed out the missing answer for non-core products. As a result the reseller offer was dropped, and the migration path, cost per investigation and shadow mode were added. The models were not used as a source of facts. - The demo uses synthetic, scripted data. It makes no network calls and connects to nothing. The MCP label in it shows how such tools could be connected; it does not describe an existing Rapid7 integration.
Market opportunity
- Securing-AI spending: $2.835B in 2026 (+83%), about $4.8B in 2027 (+68.7%), about $7.7B in 2028. Analyst estimate [19]. A narrower "agentic AI security" estimate is $1.65B in 2026, rising to $13.5B by 2032 (estimate, market-research firm) [20].
- SMBs (500 or fewer staff, n=600, fielded Aug 2025): 58% have agentic AI in production or pilot; 47% name security vulnerabilities as a top concern. Omdia survey commissioned by Dell and AMD [21].
- Coding agents: 86% of organisations deploy AI coding agents for production code (91% enterprise, 83% SMB). Survey of 500+ US technical leaders, commissioned by a coding-agent vendor, so treat with care [22]. A counterpoint: in the 2025 Stack Overflow survey, about a third of developers used AI agents [23]. McKinsey (as reported) found 22% of companies under $1B revenue use at least one AI agent [24].
- Pricing reference: public MDR list prices are rare. Huntress Managed EDR lists $8.99 per endpoint per month MSRP [25]. An aggregator lists Rapid7 Managed Threat Complete Essential on AWS Marketplace from $73,000 a year for 300 assets; I have not checked that listing directly [26].
What Kenzo does (documented) and what the demo adds
Documented: Kenzo integrates, normalises and unifies security data in an entity-centric "Security Data Mesh" (described as covering 150+ tools) [6] [17]. Specialised AI agents (threat intel, detection, Tier-2 investigation, playbook drafting) work on that mesh to investigate alerts, deploy and tune detections, hunt and prioritise response [6] [16]. Each autonomous closure carries its pivots, log excerpts and reasoning. Analysts choose where a human decision is needed ("analysts decide, AI agents do"). When an analyst marks an incident benign, a detection agent proposes suppression logic and the analyst approves it [6] [17]. Correlation covers endpoint, identity, cloud, SaaS and third-party data [6]. Reported results (customer claims, not independently checked): 94% less investigation time; alert coverage 12% → 100% [6].
Concept, not documented: treating coding agents and AI assistants as monitored entities; read-only connectors to agent vendors with fleet discovery; containment across vendors in one approval; and learning an auto-containment policy from an approval. The last one extends Kenzo's documented feedback-driven tuning (suppression proposals) to response actions. The demo labels these as concept.
Limitations
- Outside view only. I don't know Rapid7's roadmap, costs or customer data, and some of these moves may already be under way.
- Competitor capabilities come from their own public pages and were not tested.
- Figures are as of Q3 FY26.
Sources
- Rapid7 Q3 2026 financial results, 4 Nov 2026
- Q3 2026 earnings call transcript (Motley Fool)
- What Rapid7 said on its Q3 earnings call (Yahoo Finance)
- Q3 2026 earnings call highlights (GuruFocus)
- Q3 FY2026 call transcript, CEO remarks
- Rapid7 acquires Kenzo Security, 26 Mar 2026
- rapid7/rapid7-bulk-export-mcp and launch post
- Form 10-Q, quarter ended 30 Sep 2026 (SEC)
- Measuring AI Security: What MCP Exposure Really Shows (Rapid7)
- Falcon MCP (CrowdStrike)
- Introducing the Cortex MCP Server (Palo Alto Networks)
- Microsoft Sentinel MCP server, generally available (Microsoft)
- Use Tenable Hexa AI via MCP Server (Tenable)
- SentinelOne to acquire Prompt Security (agent and MCP security)
- rapid7.com homepage, read 15 Nov 2026
- Security operations is more than just investigations (Kenzo Security blog, hosted by Rapid7)
- How to evaluate agentic AI solutions to reduce risk (Kenzo Security blog, hosted by Rapid7)
- Kenzo Security raises $4.5 million (SecurityWeek)
- Gartner: securing-AI market to reach $4.8B in 2027, 26 Aug 2026
- Agentic AI security market, 2026–2032 (MarketsandMarkets, May 2026)
- Accelerating SMB growth through infrastructure and AI (Dell/AMD, Omdia survey)
- The 2026 State of AI Agents Report (Anthropic with Material)
- 2025 Stack Overflow Developer Survey: AI
- Report on McKinsey "The State of AI in 2026" (Aju Press)
- Huntress pricing
- MDR pricing: every price MDR providers publish (aggregator)