SteerPlane

SteerPlane

Runtime safety layer that stops loops, caps spend, and blocks unsafe tool calls for AI agents.

Freemium - Open-source self-hosted version is free. Cloud-hosted/managed instances and enterprise plans available by request (demo/access/book a call).

Visit Website

About SteerPlane

SteerPlane is an open-source runtime safety layer for AI agents that enforces deterministic guardrails before steps run. It stops execution loops, caps LLM spend, blocks unsafe tool calls, and applies step caps, timeouts, regex filters and cost limits. Works across agent frameworks, can be self-hosted or cloud-hosted, and provides SDK enforcement that fails closed if the control plane is unavailable. Installable via pip/npm. Patent published (v1.1.2).

Key Features

  • Deterministic pre-step checks that block unsafe calls before they reach tools
  • Stops agent loops, enforces step caps, timeouts, regex filters and cost limits
  • Caps LLM spend and prevents expensive or unsafe tool usage
  • Self-hosted or cloud-hosted deployment with SDK enforcement that fails closed
  • Installable SDKs for Python and Node.js (pip / npm)

Who is this for?

Engineering teams and enterprises building and deploying AI agents—especially startups and regulated organizations that need deterministic guardrails, cost controls and data locality.

Comments (0)

No comments yet

Be the first to share your thoughts!

Sign in to join the conversation

Sign In