Obsidian Security raised $85 million in a Series D on August 4, 2026, lifting the Palo Alto agentic security firm to a $1.1 billion valuation as enterprises race to lock down what AI agents can touch inside their SaaS estates.
Crescent Cove leads, Greylock and Menlo double down
The round was led by Crescent Cove Advisors, with participation from every existing investor including Greylock, Menlo Ventures, Norwest, IVP, Wing Ventures and Google Ventures. Obsidian said the raise follows accelerating customer growth: more than 100 accounts now spend at least $100,000 a year on the platform and 14 spend north of $1 million.
Governing what Copilot, Claude and Agentforce can do
Obsidian's pitch is that autonomous agents built on frontier models — Microsoft Copilot Studio, Anthropic Claude Code and Cowork, OpenAI, Google Vertex AI, Amazon Bedrock, Salesforce Agentforce and n8n — are gaining programmatic access to data warehouses, CRMs, dev pipelines and collaboration apps at machine speed. Obsidian discovers those agents, inventories the MCP servers and LLMs behind them, and enforces runtime guardrails on privilege escalation, excessive data access and destructive actions before impact occurs.
Why security investors like the timing
CEO Hasan Imam said more than 70% of Obsidian's customers now let agents into third-party apps. Founding Crescent Cove investor Jun Hong Heng framed the deal as a bet on foundational AI infrastructure. The round also lands as breach case studies pile up — including an agent that triggered a 13-hour cloud outage after over-broad permissions and another that deleted an entire production database — reinforcing the case that runtime governance for agents is the next control point.
For related coverage, see Glow's $180M endpoint security Series A, Anthropic's MHS agent standard, and Balance Theory's $19M Series A for CISO decision-support agents.
Reporting based on coverage from Obsidian Security, SecurityWeek, BankInfoSecurity and Yahoo Finance.