Prompt Inspector

Prompt Inspector adds prompt-attack detection and sensitive-word filtering to AI applications through an API. It is strongest for teams that want layered checks and several integration routes; buyers who need a predictable cloud price should look elsewhere.

Overview

The service checks for prompt injections, jailbreaks, and system prompt leaks. Its API-centered design suits developers adding safeguards to an existing application, while the self-hosted option gives non-commercial users another deployment route. The chief caveats are the differing Free usage limits and custom pricing for paid cloud plans.

Key features

Layered detection

Prompt Inspector combines hash caching and sensitive-word matching with semantic vector analysis, optional AI review, and arbitration. That breadth can suit teams seeking multiple kinds of checks in one service rather than relying on a single detection method. Optional review and arbitration provide additional stages, but teams will need to decide which layers suit their application.

Developer integrations and separation

Integrations include a REST API, Python and Node.js SDKs, an MCP server, and agent skills for OpenClaw and Claude Code. Cursor, VS Code Copilot, and Claude Desktop can connect through the MCP server, giving developers and users of those clients direct integration paths.

Each app has its own API key, sensitive-word configuration, rate limits, and detection logs. That separation is useful when an organization runs multiple apps and needs distinct settings and records for each.

Deployment and support

Self-hosting is available for non-commercial use under AGPL-3.0, with community support for self-hosted users. Enterprise adds private or on-premise deployment, custom integrations, managed payload library updates, and dedicated SLA support. The self-hosted license limits its fit for commercial deployments, while enterprise is the option for organizations needing private infrastructure and SLA support.

Pricing

The Free plan costs nothing and includes REST API, SDK, and MCP access. Its published allowance is 1,000 detections per month on the homepage, while the API article gives limits of 10,000 monthly requests and 10 requests per second. Since detections and requests are different measures, users should establish which cap governs their workload before depending on the free tier.

Starter, Pro, and Enterprise have custom pricing. Starter includes 100,000 requests/month, 50 req/s, and < 50 ms P95; it suits workloads within that monthly and rate ceiling. Pro raises capacity to 1,000,000 requests/month and 200 req/s, with < 30 ms P95, for higher-volume applications. Enterprise offers unlimited requests, a custom rate limit, < 20 ms P95, and SLA support for organizations needing the broadest stated capacity and service commitment. The cloud model is pay-as-you-grow, but no dollar price is provided for these paid tiers.

Platforms

Prompt Inspector supports API, Linux, macOS, web, and Windows, as well as self-hosted deployment. This range accommodates API integration across common desktop and server environments, alongside hosted access.

Who it's for

Prompt Inspector is a good fit for developers adding prompt-injection defenses to an AI application, especially when SDKs or MCP connections match their workflow. It also suits non-commercial users who can work within AGPL-3.0 and enterprises that need private deployment or dedicated SLA support. It is a weaker choice for commercial users seeking self-hosting under the stated terms, or for cloud buyers who require known prices before selecting a plan.

Pros and cons

• Pros: Multiple detection layers cover prompt attacks and sensitive words in one service.
• Pros: REST API, two SDKs, MCP, and agent skills offer several integration routes.
• Pros: Per-app keys, settings, rate limits, and logs keep application configurations separate.
• Cons: The Free allowance differs between the homepage's 1,000 detections and the API article's 10,000 monthly requests.
• Cons: Paid cloud tiers have custom pricing, making cost planning harder.
• Cons: Self-hosting is for non-commercial use under AGPL-3.0; commercial users need another arrangement.

Alternatives

For a wider set of options, browse LLM Security Tools.

• PromptGuard is worth considering for a published free allowance of 20,000 scans a month, one API key, one project, and 24-hour log retention; its Team plan is $19 a month.
• NVIDIA NeMo Guardrails is the alternative for teams wanting a free open-source Python package, though external models and services may require separate setup.
• Amazon Bedrock Guardrails may suit buyers looking for paid content filters or denied-topic controls; its stated text content filters and denied topics are $0.15 a month, while image content filters are free.
• Openlayer Guardrails is an option for teams whose needs align with its free Basic tier, which includes one member, five projects, one inference pipeline per project, 20,000 inference logs/month, 20 tests per project, one workspace, and three months of data retention.
• CyberSecEval is a free open-source benchmark suite for teams seeking a benchmarking tool.
• GLACIS is another freemium 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 developed or maintained.
• Sanctum Runtime is an alternative with a MIT-licensed repository and an open-core model.

Verdict

Choose Prompt Inspector if you are adding prompt-attack and sensitive-word checks to an application and its API, SDK, or MCP integrations fit your workflow. Its layered detection and enterprise deployment options are the main reasons to consider it. Look elsewhere if a clear cloud price is essential, if the Free cap must be unambiguous, or if you need commercial self-hosting under the stated license.