Govern what AI does in your systems
AI clients connect to an MCP Connection athttps://app.oriant.app/mcp/{name} instead of straight to the MCP server (Linear, Slack, PostHog, Cloudflare, Asana and others). For every request, Oriant:
- Signs the person in to Oriant, then to the MCP server with their own account. Their server token is stored encrypted and never handed to the AI client.
- Decides whether they may use the tool, from the MCP Policies attached to their directory groups.
- Records the MCP Call before forwarding it. A call that can’t be recorded is not sent.
See how people use AI
The endpoint agent records the prompts people send in Claude on the web, with the sender’s email. Responses are not recorded. From those prompts, AI adoption shows who uses AI, how often, in which teams, since when, and for what kind of work. The same agent scans each machine for installed AI clients and the MCP servers they’re set up to use, including servers that skip Oriant, and blocks AI tools you forbid. It also finds the Skills those AI clients load. Oriant scans every Skill and shows which are risky and which machines have them.Quickstart
Add a connection and make your first recorded call in about five minutes.
MCP Policies
Decide which tools each group may use.
Call log
See who called which tool, with which arguments, and what happened.
Endpoint agent
Find AI tools and MCP servers on your machines, and block what you forbid.
Skills scanning
See which Skills your agents load, which are risky, and who has them.
AI adoption
See who uses AI, how often, and in which teams.
Oriant MCP
Ask your own AI client about your workspace.
Roadmap
DLP and shared policies are next.