Blog
Security research, product updates, and MCP ecosystem insights.
Introducing SpiderShield v0.3: Runtime Guard for AI Agents
The first open-source runtime security layer for MCP tool calls. Policy enforcement, DLP scanning, and audit logging — all in one SDK.
Coming soonWhy MCP Servers Need Security Guardrails
MCP gives AI agents access to real-world tools. Without guardrails, a single prompt injection can lead to data exfiltration, destructive operations, or credential theft.
SpiderRating: Security Scores for 3,500+ MCP Servers
We scanned the entire MCP ecosystem and graded every server. Here's what we found — and why 40% of servers have critical security issues.
The Anatomy of a Tool Poisoning Attack
How attackers can embed malicious instructions in MCP tool descriptions to hijack agent behavior, and how SpiderShield detects it.
Building a Data Flywheel for Agent Security
How SpiderShield's local-first telemetry feeds back into better security patterns — without sending your data to the cloud.
Stay Updated
Get notified about new security research and product updates.