Agentsafe
- andyscott88/AgentSafe
AgentSafe is a MCP server that provides real-time URL trust scoring so AI agents can block phishing and malicious sites.
About
AgentSafe is an MCP server that gives coding agents a real-time URL trust score and enforcement path against phishing and malicious destinations, which matters when developers let Claude Code or Cursor browse docs, tickets, or user-supplied links without a human in the loop. You add the stdio server with a secret API key from agentsafe.app, and agent tool calls can consult trust signals before following redirects or scraping unknown hosts. It sits in Ship security because the value shows up when you are closing the gap between prototype agents and something you would ship to customers or run machine with broad network access. It does not replace full SOC workflows; it narrows the blast radius of bad URLs in agent-driven research and automation. Pair it with your existing review habits when agents fetch OAuth pages, dependency sites, or support links from strangers.
- Real-time URL trust scoring tuned for autonomous AI agents
- Blocks phishing and malicious sites before risky fetches
- stdio npm package agentsafe-mcp-server v1.0.1
- Requires AGENTSAFE_API_KEY from agentsafe.app
Agentsafe by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
claude mcp add --env AGENTSAFE_API_KEY=YOUR_AGENTSAFE_API_KEY agentsafe-mcp-server -- npx -y agentsafe-mcp-serverAdd your badge
Show developers this MCP server is listed on Skillselion. Paste this into your README.
| Package | agentsafe-mcp-server |
|---|---|
| Transport | STDIO |
| Auth | Required |
| Repository | andyscott88/AgentSafe ↗ |
What it does
Score URLs in real time inside the agent so phishing and malicious links get blocked before fetch or navigation.
Who is it for?
Best when you're shipping agents that browse the web, read tickets, or resolve arbitrary URLs during development and support.
Skip if: Closed pipelines that never fetch external URLs or teams that already enforce network egress solely at the infra layer with no agent hook.
What you get
Link opens and fetches can be checked against AgentSafe trust scores before the agent commits to a dangerous destination.
- MCP tool surface for URL trust checks via agentsafe-mcp-server
- Documented secret handling for AGENTSAFE_API_KEY
- Operational policy for blocking or escalating low-trust URLs in agent prompts
By the numbers
- Server version 1.0.1
- npm identifier agentsafe-mcp-server
- One required secret environment variable AGENTSAFE_API_KEY
Recommended MCP Servers
1Claw VaultHSM-backed vault secrets for AI agents (JIT fetch) plus prompt-injection and threat scanning.2
1passwordMCP server for 1Password service accounts — tools and resources for vaults and credentials18
402sentinel McpAssess an x402 counterparty's risk BEFORE paying: allow/review/block, scored on-chain (Base).1
A2adependencyinspector McpRemote MCP for A2A dependency inspector MCP, structured receipts, audit logs, and reviewer-ready evi
A2a Governance Bridge McpA2A Governance Bridge MCP server.A2aidentitytoll McpRemote MCP for A2A caller identity, scope policy, verdict receipts, and audit history.How it compares
URL trust MCP guardrail, not a full vulnerability scanner or code review skill.
FAQ
Who is AgentSafe for?
Developers and developers who run MCP-enabled agents that need to open or validate URLs and want phishing protection without building their own threat feed.
When should I use AgentSafe?
Use it during Ship security hardening and whenever agents follow links from users, docs, or the open web before production or high-privilege automation.
How do I add AgentSafe to my agent?
Install the npm stdio server agentsafe-mcp-server, set AGENTSAFE_API_KEY from agentsafe.app in your MCP environment, and register the server in Claude Code, Cursor, or your agent’s MCP config.