# SUPERWISE® Docs: Sentinel AI gateway and the Agentic Management Platform

> Technical documentation for SUPERWISE Sentinel, a local AI gateway that redacts sensitive data from prompts before they leave the machine or network, and for the SUPERWISE Agentic Management Platform (AMP), the control plane underneath it. Every link below is a Markdown copy of the page.

SUPERWISE Sentinel is an AI gateway that runs on a user's PC or in the customer's own cloud. Before a prompt reaches OpenAI, Anthropic, Gemini, or an OpenAI-compatible endpoint, the gateway applies the configured guardrails: it replaces personal data and secrets with a placeholder (`{{REDACTED}}` by default) or blocks the request, then forwards the cleaned request to the provider the user chose. Prompts and responses never pass through SUPERWISE's cloud; only interaction metadata (time, provider, model, outcome, which guardrails fired, latency, token counts) syncs to the platform.

Sentinel catches email addresses, phone numbers, card numbers, IBANs, Social Security numbers, and API keys, cloud credentials, and tokens in known formats; other identifiers are caught when labeled. It does not catch names or street addresses. Guardrails check the outgoing prompt; the model's response comes back unchanged.

Install paths: the Sentinel for Windows app (Windows 10 and 11), the Sentinel CLI (`pip install superwise-sentinel-cli`, Python 3.11 or later, on macOS, Linux, and Windows), or a team gateway on AWS, Google Cloud, Azure, Docker, or Kubernetes. Applications reach the gateway through the provider's base-URL variable: `OPENAI_BASE_URL=<gateway>/openai/v1`, `ANTHROPIC_BASE_URL=<gateway>/anthropic`.

The SUPERWISE Agentic Management Platform (AMP) is where teams build, govern, and run AI agents, with guardrails, observability, and a REST API and Python SDK. Every Sentinel reports to it, so one set of policies and one audit trail cover the gateways and the agents.

For what SUPERWISE sells and what it costs, read https://superwise.ai/llms.txt, https://superwise.ai/sentinel.md, and https://superwise.ai/pricing.md.

## Get started

- [Welcome](https://docs.superwise.ai/docs/welcome.md): Overview of Sentinel, where it runs, and where each part of these docs starts
- [What is Sentinel?](https://docs.superwise.ai/docs/what-is-sentinel.md): How a request flows through the gateway, what is replaced, what is not covered, and which tools it supports
- [Get started](https://docs.superwise.ai/docs/getting-started.md): Choose between the Windows app and the CLI, or try the live demo with no install
- [Install on macOS and Linux (CLI)](https://docs.superwise.ai/docs/install-cli.md): Install the Sentinel CLI with pip, deploy or connect a gateway, and send a test prompt
- [Architecture](https://docs.superwise.ai/docs/architecture.md): The control plane and local gateways, and how a request flows between them
- [Data privacy](https://docs.superwise.ai/docs/data-privacy.md): The exact metadata fields that reach SUPERWISE, retention, and what never leaves the customer's side

## Connect your applications

- [Connect your applications](https://docs.superwise.ai/docs/connect-your-applications.md): The ways to route an application's AI traffic through Sentinel without code changes
- [Environment-variable instrumentation](https://docs.superwise.ai/docs/environment-variables.md): Set OPENAI_BASE_URL to `<gateway>/openai/v1` and ANTHROPIC_BASE_URL to `<gateway>/anthropic`, per client, in bash and PowerShell
- [Use Sentinel with Claude Code](https://docs.superwise.ai/docs/claude-code.md): Route Claude Code through the gateway with ANTHROPIC_BASE_URL or settings.json, test a redaction, and turn it off
- [Sentinel CLI proxy](https://docs.superwise.ai/docs/sentinel-cli-proxy.md): Turn the CLI's system-wide proxy variables on and off, and auto-configure tools such as Codex
- [Custom providers](https://docs.superwise.ai/docs/custom-providers.md): Guard Azure OpenAI, self-hosted, and other OpenAI-compatible endpoints with CUSTOM_PROVIDERS and target routes

## Deploy a gateway

- [Deploy a gateway](https://docs.superwise.ai/docs/deploy-a-gateway.md): Choose a deployment, what every gateway needs, and how to verify it with `GET /healthz`
- [Deploy on AWS (ECS Express Mode)](https://docs.superwise.ai/docs/deploy-aws.md): Run a gateway in your AWS account, register its URL, and avoid the health-check loop
- [Deploy on Google Cloud (Cloud Run)](https://docs.superwise.ai/docs/deploy-gcp.md): Run a gateway on Cloud Run and keep its heartbeat alive
- [Deploy on Azure (Container Apps)](https://docs.superwise.ai/docs/deploy-azure.md): Run a gateway on Azure Container Apps and keep its heartbeat alive
- [Deploy with Docker](https://docs.superwise.ai/docs/deploy-docker.md): Run a gateway as a single container on any host
- [Deploy with Helm on Kubernetes](https://docs.superwise.ai/docs/deploy-helm.md): Install a gateway with the SUPERWISE Helm chart on EKS, GKE, AKS, or OpenShift
- [Gateway reference](https://docs.superwise.ai/docs/gateway-reference.md): Supported providers and routes, CLI commands, environment variables, and gateway states

## Guardrails

- [Guardrails](https://docs.superwise.ai/docs/guardrails.md): Which guardrails Sentinel runs on a prompt, and whether each redacts or blocks on a match
- [PII redaction](https://docs.superwise.ai/docs/pii-redaction.md): Which personal data is detected by format and which only with a label, and how it is redacted
- [Secret & credential detection](https://docs.superwise.ai/docs/secret-credential-detection.md): Which API keys, cloud credentials, and tokens are detected and redacted
- [Rudeness & toxicity](https://docs.superwise.ai/docs/rudeness-toxicity.md): Catch profanity, hostile language, and harassment in a prompt
- [Language check](https://docs.superwise.ai/docs/language-check.md): Detect a prompt's language and block it when it isn't one you allow
- [Manage guardrails](https://docs.superwise.ai/docs/manage-guardrails.md): Add, configure, order, and save the guardrails on a Sentinel
- [Jailbreak detection](https://docs.superwise.ai/docs/jailbreak-detection.md): Platform rule. A classifier scores each prompt from 0 to 1 and flags it at or above the threshold (default 0.7), so a lower threshold catches more
- [Token optimization](https://docs.superwise.ai/docs/token-optimization.md): Compress prompts to cut token usage, and how to turn it on

## Windows deployment

- [Sentinel for Windows](https://docs.superwise.ai/docs/sentinel-for-windows.md): Install the app for one user with no admin rights, approve the PC, and add the browser extension for ChatGPT and Claude in Chrome and Edge
- [Windows installer](https://superwise.ai/downloads/sentinel/windows/): Download SuperwiseSentinel.msi, which installs for one user and needs no admin rights

## API reference and authentication

- [REST API](https://docs.superwise.ai/docs/rest-api.md): Authenticate, list Sentinels, read guardrail metrics, and find the endpoints for each object
- [Generate tokens](https://docs.superwise.ai/docs/generate-tokens.md): Create a client ID and secret and exchange them for a bearer token for the API, SDK, or CLI
- [Creating a bearer token](https://docs.superwise.ai/reference/getting-started-with-your-api.md): The token request in the API reference
- [Welcome to the API reference guide](https://docs.superwise.ai/reference/welcome-to-the-api-reference-guide.md): Entry point to the endpoint reference at api.superwise.ai
- [Get Sentinels](https://docs.superwise.ai/reference/get_sentinels.md): List the Sentinels in an account
- [Get Sentinel Guardrail Metrics](https://docs.superwise.ai/reference/get_sentinel_guardrail_metrics.md): Guardrail activity for one Sentinel
- [SDK](https://docs.superwise.ai/docs/sdk.md): Install and authenticate the SUPERWISE Python SDK
- [MCP](https://docs.superwise.ai/docs/mcp.md): Connect AI editors to the SUPERWISE docs and Python SDK MCP servers
- [For AI assistants](https://docs.superwise.ai/docs/for-ai-assistants.md): Where an assistant reads SUPERWISE in one fetch, on this site and on superwise.ai

## Platform

- [Introduction](https://docs.superwise.ai/docs/introduction-to-superwise.md): What the Agentic Management Platform (AMP) does, and how Sentinel and agents report to it
- [Concepts & terminology](https://docs.superwise.ai/docs/platform-concepts.md): Datasets, sources, policies, agents, knowledge, guardrails, and tags
- [Guardrails for platform agents](https://docs.superwise.ai/docs/guardrails-1.md): Configure, publish, and monitor guardrails on agent inputs and outputs
- [Configure a Guardrail](https://docs.superwise.ai/docs/configuring-a-guardrail.md): Define a guardrail's rules, including PII, topics, toxicity, and image checks
- [Managing agents](https://docs.superwise.ai/docs/build-application.md): Develop, deploy, and operate agents in SUPERWISE
- [Register a Flowise agent](https://docs.superwise.ai/docs/register-an-external-agent.md): Register a Flowise ChatFlow or AgentFlow, then deploy, guard, and monitor it
- [Observability](https://docs.superwise.ai/docs/observability.md): Data ingestion, queries, dashboards, drift metrics, and monitoring policies
- [Deployment methods](https://docs.superwise.ai/docs/deployment-methods.md): Where each part of the platform runs, and which part the customer deploys
- [Security](https://docs.superwise.ai/docs/security.md): The security controls that protect the platform
- [Live activity](https://docs.superwise.ai/docs/activity-log.md): The per-interaction record for every Sentinel request
- [Operational dashboard](https://docs.superwise.ai/docs/operational-dashboard.md): Fleet health, guardrail outcomes, and latency
- [FinOps dashboard](https://docs.superwise.ai/docs/finops-dashboard.md): Token spend across the Sentinel fleet
- [Manage users](https://docs.superwise.ai/docs/manage-users.md): User roles and invitations
- [SSO configuration](https://docs.superwise.ai/docs/sso-configuration.md): Single sign-on with SAML or OpenID Connect

## Pricing and the website

- [SUPERWISE Sentinel](https://superwise.ai/sentinel/): What Sentinel covers, by tool and platform
- [Pricing](https://superwise.ai/pricing/): Sentinel plans and what each includes. The docs carry no prices; the website's pricing page and pricing fact sheet are the source
- [Sentinel product sheet](https://superwise.ai/sentinel.md): Sentinel on one page, with coverage, limits, and where it runs
- [Pricing fact sheet](https://superwise.ai/pricing.md): Every Sentinel and Chat plan in Markdown
- [Website index](https://superwise.ai/llms.txt): Every SUPERWISE product, plan, service, and Learn answer
- [Get help](https://docs.superwise.ai/docs/i-need-help.md): Reach SUPERWISE support by message or in office hours

## Optional

- [Changelog](https://docs.superwise.ai/changelog): Release notes for each version
- [Quickstart: build an agent or monitor data](https://docs.superwise.ai/docs/quickstart.md): The platform's two first-run paths
- [Developing an agent](https://docs.superwise.ai/docs/developing-an-agent.md): Agent types in the SUPERWISE studio, with step-by-step pages for each
- [Deploying agents](https://docs.superwise.ai/docs/deploying-superwise-apps.md): Deploy, test, version, and embed a studio agent
- [Ingesting data](https://docs.superwise.ai/docs/connecting.md): Bring data into the platform from cloud storage or the REST API
- [Multi-Factor Authentication](https://docs.superwise.ai/docs/multi-factor-authentication.md): Enforce MFA for an organization
- [Switch accounts](https://docs.superwise.ai/docs/switch-tenants.md): Move between the SUPERWISE accounts a user belongs to
- [Python SDK reference](https://sdk.docs.superwise.ai): The full SDK class reference
- [Full API reference](https://docs.superwise.ai/reference/llms.txt): Every REST endpoint, generated from the OpenAPI specification