Products · AgentFirewall MCP
AgentFirewall MCP
Policy firewall for agent tool calls
Allowlist, sandbox, registry-trust, prompt-injection, and redaction checks for MCP tool calls before they reach downstream servers.
Category
Agent Security
Access model
Hosted on Monarchic-managed infrastructure
Best for
Teams letting agents call tools, APIs, terminals, or internal MCPs.
What you get
- Hosted MCP route for AgentFirewall
- Scoped API keys for production agents
- Launch timing depends on operational readiness
What it does
Policy-check agent actions before they reach risky tools.
Outcomes
- Block prompt-injection-shaped tool calls before execution
- Require explicit authority for read, write, execute, and admin actions
- Redact sensitive prompt material and verify provenance claims
Primary workflows
Tool-call preflight Prompt scanning Redaction review Provenance verification
Proof boundary
Strict MCP input schemas and policy tests cover each public firewall tool.
How it fits in
AgentFirewall MCP runs as a hosted MCP route under api.monarchic.io/mcp/agentfirewall . Issue scoped API keys from the dashboard and point any MCP-aware agent client at the route when launch access opens.