Shadow AI discovery
A live inventory of AI tools and agents in use, including the ones that arrived without a ticket.
Roll out AI agents without losing control of them
IT is asked to enable Cowork, coding assistants, and a growing list of plugins across the company. Knostic gives you the inventory, the vetted catalog, and one place to set policy, so rollout is fast and still yours to govern.
Every agent is a new install surface.
Agents arrive with their own extension marketplaces, MCP connectors, and skill libraries. Each is a software supply chain IT did not choose and cannot patch. Knostic makes that supply chain visible and vetted, and gives you one control plane instead of a settings page per tool.
How Knostic helps IT
Know what is installed
Before Knostic
Users install AI tools one after another. The asset inventory does not have a column for them.
After Knostic
Kirin discovers every agent, model, and plugin on the estate, by user and team, and keeps the list current as usage changes.
Vet before install
Before Knostic
A user installs a skill from a marketplace. It carries a hidden instruction and ships a credentials file to a remote server.
After Knostic
AgentMesh gives every skill, MCP server, and extension a verdict. Dangerous ones are blocked before they reach the agent.
One settings page
Before Knostic
Claude Code, Cursor, and Copilot each expose autonomy, secrets, and connector controls differently. IT configures each one by hand.
After Knostic
Set policy once and Knostic applies it to every tool, flagging drift when a user changes a setting.
Key capabilities
Vetted catalog
AgentMesh verdicts on skills, MCP servers, and VS Code extensions, so users install from a trusted list.
Central configuration
One policy for agent autonomy, secret handling, and connectors across every supported tool.
Lightweight deployment
Kirin installs as a plugin or extension in the agents people already run. No proxies and no new network appliance.
Per-group allowances
Different policies for engineering, finance, and contractors from one console.
Reporting
Adoption, coverage, and blocked-action reports for leadership.
Frequently asked questions
As a plugin or extension inside each supported agent, pushed through your existing software distribution. There is no network appliance or proxy.
Yes, from what AgentMesh has marked clean. Dangerous and risky items are blocked or require approval, depending on your policy.
Yes, through a plugin, alongside Cursor, Claude Code, GitHub Copilot, Windsurf, Codex, JetBrains, Devin Desktop, and Gemini CLI.
Discovery shows you they exist and who uses them. You decide whether to bring them under policy, restrict them, or replace them.
Latest research and news
When “Auto-Signing” Sends Your Wallet: A Malicious MCP Server on npm
SaassyCode Repackaged: Four New VS Code Extensions and a New Delivery Chain
CodeRelay: 12 VS Code Extensions Disguised as Developer Tools
Introducing AgentMesh Access Tiers: Building Together, Growing Together
Lessons Learned from the Hugging Face Security Team
