Prompt Inspector

Prompt Inspector is an API-first safety layer for applications that handle AI prompts. It suits developers and security-conscious teams that want several detection methods, SDKs, and MCP connectivity in one service. Its deployment flexibility is a strong reason to consider it, while custom cloud pricing and differing Free quotas make it less suitable when costs must be predictable from day one.

Overview

Prompt Inspector checks incoming prompts for injection attacks, jailbreak attempts, and attempts to expose system prompts, while also filtering sensitive words. Because the checks run through an API call, it fits products that already have an application or agent workflow where a guardrail can be inserted. Teams seeking a dedicated API rather than a standalone end-user tool will find the model natural; buyers wanting a straightforward, published cloud bill will have to request pricing.

Key features

Multiple detection layers

The engine combines hash caching, sensitive-word matching, semantic vector analysis, optional AI review, and arbitration. That combination covers both fast pattern checks and more contextual analysis, giving a team several ways to tune protection instead of relying on one rule set. Optional AI review adds another decision stage for higher-risk workflows, although operating several layers requires deliberate configuration and monitoring.

Developer and agent integrations

Integration choices include a REST API, Python SDK, Node.js SDK, MCP server, and agent skills for OpenClaw and Claude Code. Cursor, VS Code Copilot, and Claude Desktop can connect through the MCP server. This breadth makes Prompt Inspector practical for conventional application back ends as well as teams working inside supported coding and assistant clients.

Application isolation

Each app receives its own API key, sensitive-word configuration, rate limits, and detection logs. That separation is useful when one organization operates multiple products or environments and needs policies and usage records kept apart, rather than sharing one global configuration.

Deployment and support

Self-hosting is offered for non-commercial use under the AGPL-3.0 license, with community support. Enterprise customers can choose private or on-premise deployment, custom integrations, managed payload-library updates, and dedicated SLA support. The licensing boundary matters: commercial self-hosting belongs in an enterprise conversation rather than the free self-hosted route.

Pricing

Prompt Inspector uses a freemium model. The Free plan costs nothing and includes REST API, SDK, and MCP access. The homepage allowance is 1,000 detections per month. An API article also describes Free limits of 10,000 monthly requests and 10 requests per second, so teams should confirm whether their workload is measured as detections or requests before committing to a production design.

Starter has custom pricing for 100,000 requests per month, a 50-request-per-second limit, and a sub-50 ms P95 target. Pro has custom pricing for 1,000,000 requests per month, 200 requests per second, and a sub-30 ms P95 target. Enterprise has custom pricing, unlimited requests, a custom rate limit, sub-20 ms P95 performance, and an SLA. Starter suits a growing service that needs a defined throughput tier; Pro is aimed at substantially busier applications; Enterprise is the choice for organizations requiring contractual support, deployment control, and the highest stated performance target.

The Free tier is a sensible evaluation path for low-volume integrations, but its monthly cap and quota terminology need clarification. Paid tiers offer progressively higher capacity and latency targets, yet none has a published amount, so budget approval requires a direct quote.

Platforms

Prompt Inspector supports API, Linux, macOS, self-hosted, web, and Windows environments. The API and SDK approach accommodates services running on common operating systems, while self-hosting provides a route for non-commercial deployments that need control over where the guardrail runs.

Who it's for

Choose Prompt Inspector for an AI application that needs prompt-injection, jailbreak, system-prompt-leak, and sensitive-word checks behind an API. It is especially suitable for teams using Python or Node, MCP clients such as Cursor or Claude Desktop, or multiple apps that need separate keys and policies. It is a weaker fit for a small project that only needs a simple open-source library, for commercial self-hosting without an enterprise arrangement, or for procurement teams that require transparent cloud prices and one unambiguous quota.

Pros and cons

Pros

• Layered protection: hash, lexical, semantic, optional AI, and arbitration stages cover different classes of prompt risk.
• Broad integration surface: REST, Python, Node, MCP, and agent skills reduce the need to build a custom connector.
• Per-app controls: separate keys, configurations, limits, and logs support multi-application operations.
• Deployment range: non-commercial self-hosting and enterprise private or on-premise options address different control requirements.

Cons

• Cloud prices are custom: buyers cannot calculate a paid-tier budget from the published plan names alone.
• Free limits need reconciliation: 1,000 detections per month on the homepage differs from the 10,000-request API allowance, with a separate 10-request-per-second figure.
• Commercial self-hosting is not the free path: the stated AGPL option is for non-commercial use, pushing commercial deployments toward Enterprise.

Alternatives

PromptGuard is the better pick when a freemium service with a clearly stated 20,000-scan monthly Free allowance, one API key, one project, and 24-hour log retention is more important than Prompt Inspector's layered options; its Team plan costs $19 a month.

NVIDIA NeMo Guardrails suits teams wanting a free, open-source Python package, with external models and services requiring separate setup.

Amazon Bedrock Guardrails is for buyers who prefer a paid web service with text content filters and denied topics at $0.15 a month in Standard and Classic tiers.

Openlayer Guardrails is an alternative for a freemium API, self-hosted, and web product with a Basic tier offering one member, five projects, and 20,000 inference logs per month.

GLACIS provides a freemium API, self-hosted, and web option with a Free account that requires no card and takes about a minute to set up.

LLM Guard is a free open-source option, but its repository is archived and no longer actively maintained.

CyberSecEval is a free open-source benchmark suite rather than a packaged guardrail service.

Sanctum Runtime is a freemium, open-core alternative documented as MIT-licensed.

For a wider comparison, see the LLM Security Tools list.

Verdict

Prompt Inspector is a strong choice for teams adding API-based prompt-safety checks to an existing application, particularly when multiple detection layers, MCP connections, and per-app isolation matter. Its main weakness is commercial uncertainty: paid prices are custom and the Free quota is expressed in conflicting units. Pick it when integration breadth and deployment options outweigh that uncertainty; look elsewhere when you need a plainly priced, single-purpose open-source guardrail or a fully predictable usage allowance.