Allow or block each AI agent tool call before it reaches your third-party apps, with every session tied to the person and tenant behind it.
Employees can connect AI agents to apps like Salesforce, Snowflake, and Slack through MCP in minutes. On a direct connection, an agent can call any tool the vendor’s server offers, including deletes and exports. The app’s access token sits in the AI client, and the app checks only that the account can act, not whether the action fits your policy.
This brief shows how the Obsidian MCP Security Proxy fixes that. It sits between AI clients and vendor MCP servers, maps each session to the person and app tenant behind it, and checks each tool call against your policies before it reaches the app. Agents get a proxy-issued token instead of the app’s own, and every decision is recorded.
Read it to see how the proxy works and how it blocks a departed employee’s agent on its next tool call.