OpenSecureAI Prompt Firewall

OpenSecureAI Prompt Firewall applies a policy to text before it reaches an LLM, deciding whether to allow, flag, or block it. It combines prompt-injection detection with secret detection and PII redaction: injection or secrets trigger a block, while PII is redacted and flagged. The live demonstration processes text in the browser and states that nothing is uploaded. Developers can also use the open-source, dependency-free engine as an npm package or CLI in applications and CI. The CLI reads files or standard input and can return a nonzero exit code when findings meet a selected threshold. The middleware can be applied to model inputs and outputs. An API key unlocks hosted detections for obfuscation or encoding evasions, indirect injection, tool abuse, multilingual overrides, and de-obfuscation. Its LLM Gateway supports OpenAI, Anthropic, xAI, and Groq with bring-your-own-key operation. The Free plan is $0/month and includes 2,000 API requests monthly; Pro is $49/month and includes 50,000 requests monthly. The product is positioned for AI application teams and developers building agents.

Prompt Firewall offers both a browser demonstration and a developer package for applying text policies before or after model use. The free plan has a monthly request limit, while additional hosted detections require an API key.

Prompt Firewall suits AI application teams shipping LLM features and developers building agents. The package and CLI may suit teams adding text checks to applications or CI.

OpenSecureAI Prompt Firewall is a policy-based screening layer for text sent to or returned by an LLM. It is best suited to developers and AI teams who want to put an open-source filter into an application or CI workflow. Its appeal is a choice of browser, self-hosted, and hosted use; the trade-off is that the free API quota is limited and several advanced detections require an API key.

Overview

The firewall runs separate prompt-injection and PII/secrets engines, then applies a user-controlled policy to classify text as allowed, flagged, or blocked. Injection and secrets are blocked, while PII is redacted and flagged. Because the middleware can process both model inputs and outputs, teams can use the same policy layer on either side of a model call.

The browser demonstration processes text entirely on the client and says nothing is uploaded. Developers can also use the dependency-free, open-source TypeScript engine through its npm package or CLI. In CI, the CLI reads files or standard input and can return a nonzero exit code when findings reach a chosen threshold—a useful way to make selected checks part of a build gate.

Key features

• Distinct responses to different risks: Blocking injection and secrets while redacting and flagging PII avoids treating every finding alike. Custom policies add room to adapt those decisions to an application’s needs.
• Local package and CI support: The open-source engine and CLI give developers a route to application-level or pipeline screening without depending on the browser demonstration. TypeScript is the listed SDK language, so teams using other languages should assess that fit before adopting it.
• Optional hosted detections: With an API key, hosted checks add coverage for obfuscation or encoding evasions, indirect injection, tool abuse, multilingual overrides, and de-obfuscation. These extend the core checks, but they make the most advanced detection capabilities dependent on a hosted service.
• LLM Gateway: The gateway supports OpenAI, Anthropic, xAI, and Groq using customers’ own provider keys. OpenSecureAI says it does not store or log those keys. That provider choice is useful for teams already using those services, but it does not remove the need to manage their own credentials.

Pricing

The Free plan costs nothing, billed $0/month. It includes in-browser tools and npm packages, 2,000 API requests per month, a limit of 30 requests per minute per key, account-synced scan history, a BYOK LLM Gateway, BYOK AI remediation, and Community support. It is a sensible starting point for evaluation or low-volume use, but the request cap and rate limit may constrain a production workflow.

Pro costs $49 a month, billed $49/month. It raises the allowance to 50,000 API requests per month and 120 requests per minute per key, and adds hosted AI remediation, LLM Gateway completions, and priority email support. Teams needing that headroom or the hosted completions have a clear upgrade path; developers who only need the open-source package may not need to pay for it.

Team has custom pricing. It includes everything in Pro, with 250,000 API requests per month and 300 requests per minute per key. Team seats and SSO, along with custom rules and an SLA, are planned. The published plan therefore has higher quotas, but those planned capabilities should not be treated as current features.

Platforms

Prompt Firewall is offered for web and self-hosted use, with hybrid deployment. The browser demo provides a client-side option, while the package and CLI support developer workflows. The core engine and npm packages are open source and dependency-free, giving teams more control over where they integrate the checks.

Who it's for

This is a strong fit for AI application teams shipping LLM features and developers building agents who want configurable text screening before or after model use. It is less compelling for teams that need broad language SDK coverage or want advanced hosted detections without relying on an API key.

Pros and cons

• Pro: One policy layer can screen both model inputs and outputs, with different actions for injection, secrets, and PII.
• Pro: The open-source, dependency-free TypeScript engine and CLI support self-hosted application and CI use.
• Pro: The client-side demo says it uploads nothing, making it useful for trying the browser workflow without sending text to a server.
• Con: Free API use is capped at 2,000 requests monthly and 30 requests per minute per key, which limits sustained use.
• Con: Hosted evasive-threat checks and Pro completions require an API key, so the full feature set is not contained in the local engine alone.
• Con: Team seats, SSO, custom rules, and an SLA are planned rather than current inclusions.

Alternatives

AI Guardrail Software is a useful category page for comparing guardrail options. Consider Enkrypt AI Guardrails if its Explore plan’s 500 starting credits, 50 credits per month, seven-day data retention, and community support better match a small trial; its Launch plan is $149 a month. Amazon Bedrock Guardrails may suit readers seeking a paid web option with priced text content filters and denied topics, each at $0.15 a month for Standard and Classic tiers; its image content filters are free.

Choose Fiddler Guardrails if its free plan’s real-time coverage for harmful exposure, hallucinations, toxicity, PII/PHI, prompt injection, and jailbreak attempts is a better fit. Openlayer Guardrails may suit teams that value its free Basic plan’s one member, five projects, one inference pipeline per project, and 20,000 inferences per month. PromptGuard is worth comparing if its free allowance of 20,000 scans per month, one API key, one project, and 24-hour log retention better fits the expected volume.

CyberSecEval is an open-source benchmark suite for readers looking for evaluation rather than this product’s middleware. NVIDIA NeMo Guardrails is an open-source Python package, with external models and services potentially requiring separate setup. Oracle Mobile Authenticator is another option to consider.

Verdict

OpenSecureAI Prompt Firewall is a good choice for developers who want an inspectable, self-hostable TypeScript policy engine and a CLI that can enforce chosen findings in CI. Its flexible placement around model calls and differentiated handling of injection, secrets, and PII are the strongest reasons to choose it. Look elsewhere if you need a larger free API allowance, broader listed SDK language support, or the advanced hosted detections without an API-key dependency.