Kodem is an application security company. It helps security and engineering teams cut long vulnerability lists down to the risks that are actually loaded and reachable in running applications, stop exploits in production, and confirm that each fix worked.
It suits organisations whose scanners produce more findings than engineers can fix, and those that now ship applications with AI agents inside them and need to see what those agents do.
At A Glance
| Vendor | Kodem |
|---|---|
| Category | Application security, runtime intelligence |
| Identities covered | Secrets, AI agents and their tools and MCP servers |
| Deployment | Cloud, on-premises and air-gapped, with an in-host sensor or sensorless |
| Agentic AI | Runtime AI Security, MCP Server, Kai |
| Certifications | SOC 2 Type II |
| Funding | A reported $40 million or more |
| Market classification | Application Security; also Cloud Security |
What Kodem Does
Scanners report what could be present in an application, not what is actually running. Kodem grades each finding by the evidence behind it, from present, to loaded, reachable, executed and finally exploitable, so teams spend their time on the last few categories. Code analysis supplies the candidate issues and evidence from inside running workloads decides which matter. Remediation is treated as unfinished until runtime confirms the fix, and the same evidence covers the application layer whether or not it contains AI agents.
Products
Runtime-Aware AI SAST And SCA
This product is for teams overwhelmed by open-source vulnerability alerts. It identifies vulnerable packages actually in use at runtime, including shadow packages loaded but missing from the manifest, and catches typosquatting, dependency confusion and compromised packages. Risk is ranked using CISA KEV, EPSS and exploit maturity together with reachability, network exposure and deployment stage. It recommends which direct dependency to upgrade to fix a transitive issue, applies a validated, reversible Dynamic Fix where no upstream patch exists, and keeps an SBOM in CycloneDX and SPDX up to date, alongside an AI-BOM and a runtime-confirmed AI-RBOM. Secret Detection, sold separately, finds exposed secrets automatically.
Application Detection & Response (ADR)
ADR protects applications already in production. It learns the normal behaviour of every package, flags deviation in real time, and intercepts an exploit the moment a vulnerable or sensitive function executes. Each exploit is traced to the exact line of code, and ADR can generate a matching WAF rule or runtime guard. Detection events stream to SIEM and SOAR tools, and webhooks can trigger incident workflows.
Runtime AI Security
Runtime AI Security discovers, governs and hardens the AI agents running in an organisation’s applications. From runtime evidence it rebuilds which agents run, which models, tools, MCP servers and guardrails they use and how they delegate, and attributes each agent action to its source. Reachable AI risk is mapped to the OWASP LLM Top 10.
Kai And The MCP Server
Kai is the agentic layer across the platform. Kai Red proves what is exploitable, Kai Blue remediates and verifies the fix, and Kai Defend protects production in real time. Together with the other products this forms the Self-Healing Application offer, one loop of finding, fixing and confirming. The MCP Server lets an AI assistant draw on the same runtime evidence an analyst sees, such as which packages are loaded and which findings are reachable.
How It Works
Kodem analyses the memory of running processes, reading the code and capabilities loaded inside them, and correlates this with execution and operating system evidence. Kernel-level signals capture load and execution events.
| Component | What it does |
|---|---|
| In-host sensor | The sensor runs alongside the workload, not inside the application process, and does not modify code or add in-process instrumentation. |
| Sensorless analysis | For restricted environments, Kodem analyses memory captured out of band and can run fully air-gapped. |
Connectivity is outbound only, and source code, secrets and raw memory contents are never sent out. Runtime intelligence supports Linux and Windows.
Identities And Systems Covered
| Secrets | Exposed secrets in code and applications. |
|---|---|
| AI agents | Agents, models, tools, MCP servers and delegation paths. |
| Workloads | Kubernetes, containers, virtual machines and hypervisors, in cloud, on-premises and regulated environments. |
| Software supply chain | Open-source packages, AI frameworks, SDKs and model files. |
Company Background
| Founded | In 2021 by Aviv Mussinger, Idan Bartura and Pavel Furman. |
|---|---|
| Leadership | Aviv Mussinger is CEO and Pavel Furman is CTO. |
| Investors | Backed by Greylock. |
Note: This overview is compiled from Kodem’s publicly available information as of 4 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.