Policy decision engine and MCP gateway for AI agents

Tool Guard Core evaluates every AI agent tool call against your policies and returns allow, deny, or escalate, with a tamper-evident audit record - before the action runs. Tool Guard Enterprise adds the MCP gateway that enforces those decisions inline: forwarding allowed calls and blocking denied ones so agents can't bypass them.

For teams running AI agents that touch money, customer data, or regulated workflows.

Evaluate, decide, audit - every tool call, before it executes.
Policy templates: tiered approval thresholds, bulk-action prevention, SQL/shell/path classifiers, and content-generation gates.
Open source under Apache 2.0 - get it at github.com/dimaggi-ai/tool-guard-core.

Tool Guard

Allow, deny, or escalate every AI agent tool call before it runs - checked against your policies, with a tamper-evident record of every decision.