- Home
- AI Guardrail Software
- OpenGuardrails
On the recordRead 4 Oct 202613 facts
OpenGuardrails
Sheridan, Wyoming, United States
Plans and prices
Starter
$10a month
Secret masking on your host · Secret ledger · Detects secrets, PII, dangerous commands and credential files
Read 4 Oct 2026 · openguardrails.com
Pro
$20a month
Includes Starter · Model judges on every tool call, prompt and reply · Sized for one agent in normal use · Four policy templates
Read 4 Oct 2026 · openguardrails.com
Max
$100a month
From $100/month · $100/month for 5x or $200/month for 20x · Multimodal storage and display · Priority evaluation
Read 4 Oct 2026 · openguardrails.com
Enterprise
Priced on request
Private deployment · Events stay in your infrastructure · Enterprise SSO or admin provisioning · Support with an SLA
Read 4 Oct 2026 · openguardrails.com
On the record
- Console
- The console includes a live monitor, agent and session explorer, findings triage, policies, a playground and API keys.openguardrails.com · read 4 Oct 2026
- Integration method
- There is no SDK; integrations use the API directly, with two POSTs per model call in the documented recipe.openguardrails.com · read 4 Oct 2026
- Integrations
- The plugin index lists Higress, DeepSeek Harness, litellm, Claude Code, Codex, opencode, OpenClaw, Hermes, LangGraph and a mitmproxy gateway example.openguardrails.com · read 4 Oct 2026
- Open protocol
- The OGR protocol is an open Apache-2.0 specification with published JSON Schemas; the runtime product is commercial.openguardrails.com · read 4 Oct 2026
- Operational limit
- The default rate limit is 600 requests per minute per API key.openguardrails.com · read 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.openguardrails.com · read 4 Oct 2026
- Purpose
- OpenGuardrails AIRS tests, inspects and protects AI agents at runtime.openguardrails.com · read 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.openguardrails.com · read 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.openguardrails.com · read 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.openguardrails.com · read 4 Oct 2026
- Support
- The Enterprise plan includes support with an SLA.openguardrails.com · read 4 Oct 2026
- Verdicts
- The runtime returns allow, block or redact decisions, along with findings and redaction spans.openguardrails.com · read 4 Oct 2026
