CheckOps
Salus
Salus is a runtime authorisation layer for AI agents that checks consequential tool calls against policy, session evidence and intent before execution, and returns exact corrections to the agent.
- SaaS
- Paid
- LLM Guardrails
- Ai Security
- AI-enabled testing
- Production monitoring & observability
- Security testing
Latest public signals
Not refreshed yet. Refresh reads the public site and, when the entry has a GitHub repository, the latest release.
Introduction
Salus sits between an agent and the actions it takes. It checks policy, evidence and intent at runtime, returns exact fixes to the agent and stops wrong actions before they become real.
Its demo shows an agent about to book a $299 reservation from a stale itinerary. Salus holds the action because live availability was never verified, returns the missing step, and the agent searches live inventory, finds the new $375 price and completes the booking correctly. Plans include unlimited seats with allowances of registered agents and monthly action checks.
Launch and background: Y Combinator’s company directory lists Salus in its Winter 2026 batch, with the one-liner “Guardrails to validate your agent’s actions before they execute”.
Features
- Pre-execution checks: Validate consequential tool calls.
- Policy and evidence: Rules plus session context.
- Exact corrections: Tell the agent what is missing.
- Sandbox: Try policies safely.
- Unlimited seats: Plans priced by agents and checks.
Pros
- Prevents real damage: Stops wrong actions, not just bad text.
Cons
- Price: Plans start at $500 a month.
- Integration work: Tool calls must route through Salus.
Usage and Links (Example)
- Register your agent and its consequential tools.
- Write policies for evidence each action needs.
- Test in the sandbox, then enforce in production.
Resources:
Pricing
Salus lists plans at $500 a month and $2,500 a month with unlimited seats; the Pilot plan is month-to-month.
Recommended for?
- Teams letting agents book, pay or change records.
Conclusion
Salus is a guardrail for actions rather than words, which matters as agents gain tool access.
Treat it as a production control alongside pre-release evals.
On CheckOps, Salus is filed under the Production monitoring & observability and Security testing stages. Related tools listed here include Guardrails AI, NeMo Guardrails, Lakera, Open Policy Agent, Promptfoo and Giskard. When comparing them, weigh the licence, the languages and platforms each supports, whether it is self-hosted or managed, and how cleanly it slots into your existing CI pipeline. Details on this page were checked against the official website, documentation and repository on 11 November 2026.
Alternatives
Other tools on CheckOps that cover similar ground. Open a page, or compare one with Salus side by side.
GGuardrails AIGuardrails AI is an Apache-2.0 Python framework for adding input and output validation to LLM applications, with a hub of validators and structured output generation.Compare with Salus
NNeMo GuardrailsNeMo Guardrails is an Apache-2.0 toolkit from NVIDIA for adding programmable guardrails to LLM-based conversational systems, using Colang flows and input, output, dialog and retrieval rails.Compare with Salus
LLakeraLakera, now part of Check Point, offers AI security products: a Guard API (documented as Check Point AI Guardrails) for runtime screening of prompt injection and data leaks, and AI red teaming.Compare with Salus
OOpen Policy AgentOpen Policy Agent (OPA) is an Apache-2.0, CNCF-graduated general-purpose policy engine that evaluates Rego policies for Kubernetes, CI pipelines, APIs, Terraform and more.Compare with Salus
PPromptfooPromptfoo is an open-source CLI for LLM evals and red teaming. The community tool is free; enterprise features are quoted.Compare with Salus
GGiskardGiskard offers an Apache-2.0 open-source library for evals, red teaming and test generation for AI agents, plus a commercial Hub for agent red-teaming, evaluation and guardrails.Compare with Salus