OpenGuardrails
OpenGuardrails AIRS is a commercial runtime service that checks AI agents before they receive requests and before they act on responses.
What OpenGuardrails is
OpenGuardrails AIRS is a commercial runtime service that checks AI agents before they receive requests and before they act on responses. It evaluates tool calls, resources, and accounts before a model sees a request, then examines responses before the agent acts. The runtime can allow, block, or redact, returning findings and redaction spans. Its console includes a live monitor, agent and session explorer, findings triage, policies, a playground, and API keys. Integrations use the API directly rather than an SDK; the documented recipe makes two POSTs per model call. Starter masks secrets on the customer host before requests leave and restores them only into tool arguments. Private deployments run on customer Kubernetes or VMs, with customer datastores and a model gateway in the customer environment. Starter costs $10 a month, billed monthly with cancellation anytime; there is no free plan. The default rate limit is 600 requests per minute per API key. If the runtime is unreachable, local preconfigured fail policies default to open unless categories are explicitly set to closed. The OGR protocol is an open Apache-2.0 specification, while the runtime product is commercial.
Good
- Checks tool calls, resources, accounts, and responses.
- Can allow, block, or redact with findings.
- Starter masks secrets on the customer host.
- Private deployment keeps events in customer infrastructure on Enterprise.
Watch out
- No free plan.
- Starter costs $10 a month, billed monthly.
- No SDK; integrations use the API directly.
- Default limit is 600 requests per minute per API key.
Who it suits
It suits teams that want runtime checks for AI agents, including controls for secrets, PII, and dangerous commands. Teams considering private deployment can run it in their own Kubernetes or VM environment, while those integrating directly need to use its API rather than an SDK.
Verdict
OpenGuardrails AIRS provides runtime inspection with configurable decisions and customer-host secret masking. The API-based integration approach, monthly starting price, and default rate limit are relevant constraints when assessing fit.
Plans and prices
From OpenGuardrails's own pricing page, checked 4 Oct 2026.
| Plan | Price | What it gives you |
|---|---|---|
| Starter | $10 a month | Secret masking on your host · Secret ledger · Detects secrets, PII, dangerous commands and credential files · Billed monthly; cancel anytime. |
| Pro | $20 a month | Includes Starter · Model judges on every tool call, prompt and reply · Sized for one agent in normal use · Four policy templates · Billed monthly; cancel anytime. |
| Max | $100 a month | From $100/month · $100/month for 5x or $200/month for 20x · Multimodal storage and display · Priority evaluation · Billed monthly; cancel anytime. |
| Enterprise | Priced on request | Private deployment · Events stay in your infrastructure · Enterprise SSO or admin provisioning · Support with an SLA · Annual license; priced on fleet size. |
The facts
Each one read from OpenGuardrails's own pages, with the date it was checked.
What it does
- Console
- The console includes a live monitor, agent and session explorer, findings triage, policies, a playground and API keys.checked 4 Oct 2026
- Integration method
- There is no SDK; integrations use the API directly, with two POSTs per model call in the documented recipe.checked 4 Oct 2026
- Integrations
- The plugin index lists Higress, DeepSeek Harness, litellm, Claude Code, Codex, opencode, OpenClaw, Hermes, LangGraph and a mitmproxy gateway example.checked 4 Oct 2026
- Open protocol
- The OGR protocol is an open Apache-2.0 specification with published JSON Schemas; the runtime product is commercial.checked 4 Oct 2026
- Operational limit
- The default rate limit is 600 requests per minute per API key.checked 4 Oct 2026
- Private deployment
- Private deployments run on customer Kubernetes or VMs with customer datastores and a model gateway hosted in the customer environment.checked 4 Oct 2026
- Purpose
- OpenGuardrails AIRS tests, inspects and protects AI agents at runtime.checked 4 Oct 2026
- Runtime checks
- It evaluates tool calls, resources and accounts before the model sees a request and checks responses before the agent acts on them.checked 4 Oct 2026
- Security and privacy
- Starter masks secrets on the customer host before a request leaves it and restores them only into tool arguments.checked 4 Oct 2026
- Service behavior
- When the runtime is unreachable, integrations use local preconfigured fail policies that default to open unless categories are explicitly set to closed.checked 4 Oct 2026
- Support
- The Enterprise plan includes support with an SLA.checked 4 Oct 2026
- Verdicts
- The runtime returns allow, block or redact decisions, along with findings and redaction spans.checked 4 Oct 2026
The company
- Headquarters
- Sheridan, Wyoming, United Stateschecked 28 Sep 2026
Alternatives to OpenGuardrails
- 1
PromptGuard
Free planWebAPIExtensionLinux7.9From $19 a month - 2
CyberSecEval
Free planOpen sourceWebAPILinuxmacOS7.5Free plan - 3
NVIDIA NeMo Guardrails
Free planOpen sourceAPILinuxmacOSSelf-hosted7.4Free plan - 4
Oracle Mobile Authenticator
Free planAndroidiOSWindows6.9Free plan - 5
Enkrypt AI Guardrails
Free planWebAPISelf-hosted6.8From $1,608 a year - 6
OpenSecureAI Prompt Firewall
Free planWebSelf-hosted6.7From $49 a month