StrictBots.com — Deterministic Policy-Gated Runtime
Strict bots do not decide; they evaluate. Every action passes a deterministic policy chain — first match wins, unmatched actions deny. This page documents the runtime contract and simulates the exact evaluation in your browser.
Deterministic Policy Allow/Deny Simulator
Define allow/deny rules and fire a sample action through them. First-match evaluation, deterministic verdict — the strict-bots runtime gate, simulated entirely in the browser.
Set policies and an action, then Evaluate.
Policy semantics
- Rules evaluate top-to-bottom; first match wins — order is the security boundary.
- Tool name must match exactly; the optional argument supports
*globs. - No match means DENY — the runtime is fail-closed by construction.
- Verdicts are deterministic: same policies + same action = same verdict, always auditable.
- Denials should carry a reason code back to the agent for self-correction loops.
Runtime contract
A strict bot runtime wraps every tool invocation in this gate. Policies are data, not code — reviewable, diffable, and hot-reloadable without redeploying the agent. Pair with typedbots.dev contracts upstream so only well-formed calls reach the gate.
Cluster
Sibling properties in the MCP/A2A protocol bridge and guardrail runtimes.
MCP Context ProMCP↔A2A HubA2A→MCP CompilerOverwatcherTypedBotsStrictBotsRunWeights · ForwardRouterReverseHarnessLinkProducers