Knostic secures AI coding assistants and agents on developer machines. Its software sits between an agent and the tools, extensions and MCP servers it connects to, and stops unsafe connections or leaked secrets before the agent acts. It is built for security teams that have to let developers use tools such as Cursor and Claude Code without losing control of what those agents can reach.
The underlying problem is that coding agents act with the developer’s own access, so a single poisoned component can reach source code, credentials and internal systems.
At A Glance
| Vendor | Knostic |
|---|---|
| Category | Security for AI coding agents and the agentic endpoint |
| Identities covered | AI agents and coding assistants, MCP servers, developer endpoints |
| Deployment | User-space endpoint client, deployed through MDM or installed by developers |
| Agentic AI | MCP inspection, agent skill scanning, inline policy enforcement for coding agents |
| Certifications | SOC 2 and penetration test reports through its trust centre |
| Founded | 2023 |
| Market classification | Agentic AI Identity; also Authorisation |
What Knostic Does
A coding agent can install an extension, call an MCP server or paste an API key into a prompt, and a poisoned skill or malicious server turns that convenience into a supply-chain attack. Knostic addresses this at the endpoint, where the agent runs. It pairs enforcement on the developer machine with intelligence on which skills, servers and extensions are safe, and it carries the controls a company already has, such as DLP, DSPM, Microsoft Purview and SIEM, into AI workflows so they keep working when developers use agents.
Products
Kirin
Kirin is endpoint security for AI coding agents. It inspects MCP connections in real time, monitors IDE extensions and plugins, and detects secrets such as API keys, tokens and passwords in user prompts. Security teams control it through detection policies, AI agent instructions and allow and block lists, with defaults switched on from installation. Larger teams get role-based access control, audit logs, policy groups and SSO.
AgentMesh
AgentMesh is threat intelligence for the agentic ecosystem, for teams that need to know whether a skill or server is safe before allowing it. It continuously discovers, tracks and scans AI agent skills, MCP servers and VS Code extensions, checking them for prompt injection and supply-chain threats, and marks each one dangerous, risky or clean with the findings behind the verdict. Verdicts can be looked up by SHA-256 hash or pulled through a REST API, and the catalogue is included with Kirin in the paid AgentMesh plans.
OpenAnt
OpenAnt is Knostic’s open-source, LLM-powered vulnerability discovery tool for application security teams. It maps each function with its dependencies, callers and call context, then tries to attack it, an approach that is reported to remove up to 99.98% of false positives on popular open-source projects. It supports Go, Python, JavaScript, TypeScript, C and C++, PHP and Ruby, and is also offered as a managed service.
How It Works
| Component | What it does |
|---|---|
| Endpoint client | Kirin runs in user-space on the developer machine and does not need admin rights. It can be pushed through an existing MDM with scripts Knostic provides, or installed directly by developers. |
| Inline enforcement | Kirin checks dependencies, extensions and MCP connections against policy before the agent acts. |
| Intelligence | AgentMesh verdicts are available in a web interface and through the REST API. |
| Management | Team and Enterprise editions manage policy for the whole organisation from one dashboard and stream events to the customer’s SIEM. |
Identities And Systems Covered
| AI coding agents | Cursor, Claude Code, GitHub Copilot, Windsurf, Codex, JetBrains, Devin Desktop and Gemini CLI |
|---|---|
| MCP servers and skills | MCP connections inspected by Kirin and catalogued with skills by AgentMesh |
| Extensions | IDE extensions, plugins and VS Code extensions |
| Secrets | API keys, tokens and passwords in prompts |
Company Background
| Founded | 2023 |
|---|---|
| Headquarters | Herndon, Virginia, United States |
| Leadership | Sounil Yu (Co-Founder and CTO) |
| Funding | An $11 million round led by Bright Pixel Capital in March 2025, after a $3.3 million pre-seed in April 2024, for a reported $19 million in total; other investors include DNX, Pitango, Seedcamp, Shield and SVCI |
| Recognition | Included in the RSAC Innovation Sandbox, Gartner Cool Vendors and the SINET16 |
Note: This overview is compiled from Knostic’s publicly available information as of 3 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.