P

LLM Security Tools

PromptGuard: AI Request Security and Plans

promptguard.co · checked 14 Oct 2026

PromptGuard checks app requests for security risks before they reach an AI model.

How it plugs in

—Public API
NoOpen source
YesFree tier

Signal on LLM Security Tools6.7#4 of 31

Positioned in the path from an application to its LLM provider, PromptGuard screens requests for issues including prompt injection, information exposure, and risky tool calls, and supports personal-data redaction. It works with several model providers and offers cloud and self-hosted deployment. Teams can start on its free tier or choose paid options for broader use.

PromptGuard home page

Review

In a line

Before an app passes a request to an AI model, PromptGuard examines it as a security layer. It aims to block threats and protect sensitive information.

What it does

PromptGuard checks for risks such as prompt injection, jailbreak attempts, data exfiltration, and malicious tool calls. Its 15 detectors are arranged in six layers and also cover items such as secrets, malware, and unsafe content.

Its personal-information controls detect and redact 43 types of data using reversible tokenization, which replaces sensitive text with tokens that can later be restored. It also offers automated adversarial testing across more than 20 attack types.

In PromptGuard's published evaluation, precision was 99.1%, with 10 benign samples flagged out of 991; recall was 80.3%, with 271 attacks missed among 1,378 tested. These figures describe its published test results, not a guarantee for every use.

How it connects

PromptGuard lists compatibility with providers and platforms such as OpenAI, Anthropic, Gemini, Azure, Bedrock, Mistral, Cohere, DeepSeek, Groq, HuggingFace, Ollama, and vLLM.

People can connect it through an SDK that can automatically cover supported model calls, a direct API, an HTTP proxy, or a GitHub scanner. The Python SDK is open source under the MIT license and supports several frameworks, including LangChain and LlamaIndex.

PromptGuard runs as a managed cloud service, in a customer's virtual private cloud, or in an air-gapped installation that has no network connection. Its hosted deployment is in Google Cloud's us-central1; a hosted EU location is not offered at present.

Good fit

PromptGuard may suit teams that want a security check between their apps and model providers, especially when they need personal-information redaction or controls for AI-agent tool calls. Deployment is available as managed cloud, self-hosting, or an air-gapped installation.

Watch for

Five of the OWASP LLM Top 10 risks are covered fully, while five have partial coverage. Gaps named include provenance verification, poisoning, vector-store isolation, and spending caps.

There is no EU-based hosting region. PromptGuard is not certified for SOC 2 Type II, ISO 27001, or FedRAMP, and does not sign BAAs for hosted HIPAA use. Its 99.9% availability figure is an internal target rather than a signed self-serve SLA.

In zero-retention mode, prompt and response bodies are not written. By default, security-event records include a preview capped at 500 characters and a content hash. The SDK defaults to allowing requests through if its Guard API is unavailable, though it can be set to block instead.

Pricing

The Free tier has no charge and allows 20,000 scans per month. Requests pause when its quota is reached. Team costs $19 a month per seat, with pooled scans and desktop and browser tools.

Build and Pro are each priced at $99 per month. Build includes 100,000 scans a month and custom policies; Pro includes the same monthly scan allowance and seven-day log retention. Enterprise pricing is available on request. Scale lists 500,000 requests a month, but no plan price is listed.

Free remains available indefinitely rather than ending as a trial. Paid plans include a 14-day money-back guarantee.

Verdict

PromptGuard is worth considering when requests need screening before they reach a model and you want the choice of managed or self-hosted deployment. Its provider compatibility and data-redaction controls may be useful for teams handling sensitive prompts. The free plan provides a way to start without a subscription, while paid plans add capacity and team features.

Look closely at the security gaps and hosting details before choosing it for regulated or EU-based workloads. The hosted service is not available in the EU, and PromptGuard does not have SOC 2 Type II or ISO 27001 certification or sign hosted HIPAA BAAs. Also check which plan's scan limits and retention period fit your needs.

Our take

PromptGuard may suit applications that need request screening before sending traffic to an AI model, especially where personal-data redaction or agent tool checks are priorities. Its provider range and deployment choices give teams flexibility in how they connect it. The Free plan offers a no-cost starting option, while paid plans add scan capacity and team features. It may not meet requirements for EU-hosted residency, SOC 2 or ISO 27001 certification, or a hosted HIPAA BAA.

  • Screens requests for prompt injection, data exposure, and related risks.
  • Detects and redacts 43 types of personal information.
  • Supports a range of model providers and frameworks.
  • Available as managed cloud, hybrid self-hosting, or air-gapped deployment.
  • The no-cost tier allows 20,000 scans per month.
  • A hosted EU region is not available.
  • PromptGuard is not certified for SOC 2 Type II, ISO 27001, or FedRAMP.
  • Hosted HIPAA BAAs are not offered.
  • Five OWASP LLM Top 10 risks have partial rather than full coverage.
  • Pick it ifTeams adding security checks between apps and AI model providers; Organizations that need a self-hosted or air-gapped option; Teams seeking personal-information redaction
  • Skip it ifYou need EU hosting, SOC 2 Type II, ISO 27001, or FedRAMP certification.; You need a hosted HIPAA BAA or complete coverage of the OWASP LLM Top 10.

Pricing

Free is a permanent plan with 20,000 scans monthly. Team is $19 per seat per month; Build and Pro are $99 monthly apiece. Enterprise pricing is custom, while Scale has no listed price. Paid plans come with a 14-day money-back guarantee.

  • Freeper month · 20,000 scans/month · 1 API key · 1 project · 24-hour logs · requests pause at quotaFree
  • Teamper seat/month · 15,000 scans per seat pooled · browser extension · macOS/Windows agent · fleet enrollment · MDM · on-device secret blocking and PII redaction$19/mo
  • Buildmonthly · 100,000 scans/month · 5 API keys · 5 projects · full detector set · custom policies · email alerts$99/mo
  • Pro100,000 scans a month · 5 API keys · 5 projects · 7-day log retention$99/mo
  • Enterprisecustom contract · custom volume · air-gapped deployment · SCIM · IP allowlist · custom retention · signed bundles · dedicated CSM · custom SLAOn request
  • Scale500,000 requests/month · unlimited API keys and projects · 30-day log retention · overage metered at $0.40 per 1,000 requests up to spend capOn request

As promptguard.co lists them, checked 14 Oct 2026.

Questions

Is there a free PromptGuard plan?

Yes. Free includes 20,000 scans per month, one API key, one project, and 24-hour logs. Requests stop once the monthly allowance is used.

Which AI providers does PromptGuard work with?

Its listed integrations include OpenAI, Anthropic, Gemini, Azure, Bedrock, Mistral, Cohere, DeepSeek, Groq, HuggingFace, Ollama, and vLLM.

Can PromptGuard run outside its managed cloud?

Yes. Options include hybrid deployment in a customer's virtual private cloud and a fully air-gapped installation, as well as managed cloud.

Does PromptGuard offer a trial?

The Free tier is ongoing, not time-limited. Paid plans include a 14-day money-back guarantee.

Is PromptGuard hosted in the EU?

No. Its hosted service runs on Google Cloud in us-central1, and it does not currently provide an EU hosting region.

What security certifications does PromptGuard have?

PromptGuard is not currently SOC 2 Type II certified; ISO 27001 is planned. It is also not FedRAMP certified and does not sign hosted HIPAA BAAs.