Obsidian launches runtime policy engine for AI agent actions
Original title: Obsidian expands AI agent security with runtime policy enforcement
Policies can allow, block, or require approval, including for OpenAI Codex and MCP tool calls.
Obsidian Security announced a Runtime Policy Decision Engine that evaluates AI agent actions before they reach applications, using identity, access, and customer labels. Enforcement points can allow, block, or require human approval where supported. New pieces include an MCP Security Proxy and a Policy Decision API for customer gateways. Coverage spans coding agents and in-app agents in several enterprise platforms.
- Coding-agent policy now includes OpenAI Codex with Claude Code, Cowork, and Microsoft Copilot.
- The MCP Security Proxy checks tool calls and keeps app tokens out of the AI client.
- Discovery covers agents in Salesforce Agentforce, Databricks, Snowflake, ServiceNow, and Workday.
- Obsidian is partnering with Cloudflare’s MCP Server Portal for downstream access.
- Vendors
- Obsidian Security
- Organizations
- Cloudflare
From the original article
5 min read at helpnetsecurity.comObsidian Security has announced its Runtime Policy Decision Engine, giving security teams centralized control over AI agent actions before they run. Building on Obsidian’s extensive AI agent discovery and governance coverage, the engine uses context about who is behind an agent, its access and the third-party applications involved to allow or block actions, or require human approval where supported.
By Industry News · helpnetsecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.