Seismic secures its growing Claude and agent footprint with Reco

The challenge

Seismic uses Claude for its most demanding work. Adoption moved quickly across a globally distributed technical workforce, and the number of agents in the environment kept rising week over week.

The Information Security team needed to know which agents and skills existed, what data and systems they could reach, and which carried real risk. Seismic was also concerned about skills and connectors pulled from outside sources, which can carry hidden instructions or malicious code. Existing tools didn't give them a single view of this across platforms. As the Seismic team put it, "once you have a list of 1,400 agents, you're bombarded with thousands of issues and it's not clear what to prioritize." 

With Reco, Seismic:

• Connected Claude, Copilot Studio, ChatGPT Enterprise, and its other AI platforms to Reco for one inventory of agents, skills, and MCP connections
• Adopted Reco’s Agent Governance, Agent Security, and Agent Runtime solutions
• Deployed Reco's Claude Desktop collector to managed Macs through Jamf, giving visibility into Claude skills and MCP servers on employee devices
• Used Reco's risk scoring (likelihood, impact, and guardrails) to rank agents and focus on the highest-risk ones first
• Adopted Reco’s MCP server 

The results

• A single inventory of 1,400 agents across Claude and other platforms
• All high agent risks cleared
• Claude skills and MCP servers on managed devices now visible to security across 2,000 users
• Security keeps pace as adoption grows, rather than blocking it

‍

"With Claude and other AI tools spreading across our teams, we needed to see every agent and skill, and know which ones mattered. Reco gave us that view as well as a way to act on it."

Bill Young, Director Information Security at Seismic

‍