AI Discovery finds every AI tool on every machine in your company, from coding agents and AI editors to desktop assistants, AI browsers, local models and MCP servers. It shows what each one can do and reach, and lets you sanction or block it for the whole team.
ChatGPT on the desktop, Ollama serving models, an AI browser, a sidebar extension in Chrome. None of it crosses a proxy. The only way to know it exists is to look at the machine.
A sanctioned agent is still a risk when it skips permission checks, runs with its sandbox off, or runs from a scheduled job with no one at the keyboard.
MCP servers wired into production databases, credentials in home folders, packages fetched unpinned on every start. No one has written down what each agent can reach.
A catalog of AI tools in every category, from coding agents and AI editors to desktop assistants, AI browsers, local model runtimes, SDKs, browser extensions and MCP servers. Tools the catalog has never seen still show up: model files, apps with an AI runtime inside, extensions that declare AI features.
One tool, across the fleet: who runs it, how it signs in, what it can do, what it reaches.
An exfiltration path has three legs: the agent reads private data, takes in content outsiders can write, and can send data out. AI Discovery assembles each leg from the tool's capabilities, its MCP servers, the credentials in reach and its live connections, then asks who approves.
Risks: a complete path with no one approving is critical. The same tool with a person at the keyboard is not.
Every newly found tool waits as Unsanctioned. Allow or block it once, with a note, and the decision applies to every endpoint on the team. AI Discovery supplies the facts; Orpheus enforces.
Policy: one decision per tool, team-wide, with who decided and why.
One line from the Orpheus install page, on macOS, Windows or Linux. The first scan runs right away; the scanner keeps itself scheduled and up to date.
Every hour, in the background, done in seconds. It reads what is installed and how it is set up; never files, prompts or secrets.
Open Orpheus to see the directory, the risks and the policy for the whole team. Decide once, and every machine follows.
Cassandra is the free, one-machine cut of this: it checks your extensions against 11,174 known-malicious ones, once, on the machine you run it on. AI Discovery is what a team needs after that.
Cassandra Free
AI Discovery Part of Orpheus
Book a 30-minute walkthrough. The first scan takes seconds; the directory fills in as machines report.
