CodeIntegrity
Runtime guardrail platform for production AI agents, built by offensive-security researchers who publicly compromised the Notion app in under four hours.
Visit Website ↗ + Add to CompareOverview
CodeIntegrity builds runtime guardrails for AI agents deployed in production, aiming to give engineering and security teams a way to constrain what an agent can actually do at execution time rather than relying solely on prompt-level instructions or after-the-fact monitoring. The company positions its approach as protection that persists even when an agent is manipulated or behaves unpredictably, rather than protection that depends on the agent behaving as instructed.
Founded in San Francisco in 2024 by CEO Steven Jung and CTO Abi Raghuram, CodeIntegrity built its early reputation on offensive security research, including a widely cited exercise in which the founders compromised the Notion application in under four hours — work referenced by The Economist in 2025. The company announced a $5 million seed round in May 2026, led by SYN Ventures with continued participation from Antler and Boost VC.
CodeIntegrity’s offensive research background gives it more independently notable technical credibility than many AI-agent security peers at a similar funding stage, even though its guardrail platform itself has not yet undergone public independent testing.
Innovation Matrix Assessment
Moved from offensive-research notoriety (the Notion compromise) to a funded, productized runtime guardrail platform within about two years of founding.
Runtime guardrails for production AI agents address a real operational gap for engineering teams shipping agentic features faster than they can secure them.
A $5M seed from SYN Ventures, Antler, and Boost VC is a modest but credible early round for a two-year-old company.
Runtime guardrails for agentic AI are a needed evolution of application security, but CodeIntegrity competes with a fast-growing field of similarly positioned AI-agent security startups.
The founders' Notion compromise is independently notable, credible offensive-research evidence of skill, though it demonstrates the team's capability rather than independently validated efficacy of the CodeIntegrity product itself.
As more production applications embed AI agents, runtime-level protection against unpredictable agent behavior is likely to stay a persistent need.
Why CISOs Should Care
It gives application security teams a way to put hard limits on what a production AI agent can do, even if the agent itself is manipulated or misbehaves.
What Makes It Different
Its guardrails are enforced at runtime and hold even when an agent is compromised, rather than relying on the agent following its instructions correctly.
The Matrix Verdict
55/100 — INCREMENTAL INNOVATOR
CodeIntegrity is an Incremental Innovator: genuinely credible offensive-security pedigree (the Notion compromise, cited by The Economist) applied to a real and growing problem, still awaiting independent validation of the product itself.
Editorial Note: Claims vs. Verified Findings
Founders, funding, and the Notion compromise research are independently reported (GeekWire, The Economist reference, TheSaaSNews). Specific guardrail efficacy claims for the commercial product are vendor-sourced and not yet independently tested.
Sources
Alternatives to CodeIntegrity
Endor Labs
Reachability-based software composition analysis that filters open-source dependency risk down to what code paths are actually exploitable.
ThreatLocker
CISO ReviewedZero Trust endpoint protection platform that blocks unknown applications by default through allowlisting, ringfencing, and storage control.
Socket
A software supply chain security platform that combines automated dependency analysis with human verification to catch malicious and…
JFrog
Public software supply chain security platform (Xray plus Advanced Security) that scans binaries and artifacts end-to-end from the…
Apiiro
Builds a continuous 'code risk graph' that maps code, developers, and cloud deployment to prioritize AppSec findings by…
ArmorCode
Tool-agnostic ASPM layer that correlates findings from 300+ existing security tools into one prioritized backlog, without running its…