# PromptGlass

> PromptGlass is an AI usage monitoring and AI governance platform for organizations. It shows which employees and teams use which AI tools (ChatGPT, Claude, Microsoft 365 Copilot, GitHub Copilot, Gemini, Cursor, model APIs and local LLMs), captures prompts and responses under configurable privacy controls, detects and prevents sensitive data leakage (AI DLP), and measures AI adoption, efficiency and cost. It runs as SaaS (US, EU, Canada regions), self-hosted on Kubernetes via Helm, or air-gapped on-premises.

## Key facts
- Capture: four sensors. (1) Gateway: forward proxy that inspects only AI-catalog domains, plus an OpenAI- and Anthropic-compatible API gateway (set OPENAI_BASE_URL / ANTHROPIC_BASE_URL). (2) Browser extension for Chrome, Edge, Firefox, Safari. (3) Endpoint agent for Windows, macOS, Linux. (4) Read-only SaaS API connectors (Microsoft 365 Copilot via Microsoft Graph, OpenAI, Anthropic, GitHub Copilot, Google Workspace, Azure OpenAI, Amazon Bedrock).
- Identity: SSO and SCIM from Entra ID, Okta or Google Workspace; usage attributed to person, department and cost center.
- Detection: PII, PHI, PCI card numbers (Luhn-validated), national IDs, secrets and API keys, source code, custom regex/dictionaries, exact data match.
- Policy actions: allow, log, coach, redact (with optional restore in the response), block, reroute to an approved model. Policies are YAML, versioned and testable by replay simulation.
- Analytics: adoption by team and app, use-case clustering, prompt quality and re-prompt rates, seat utilization (idle ChatGPT/Copilot seats), duplicate tools, token cost by team and model, model right-sizing.
- Privacy: pseudonymized users by default, storage modes (full, redacted, metadata-only, hashed), four-eyes re-identification, customer-managed keys, regional data residency. Customer prompts are never used to train models.
- Self-hosted stack: Kubernetes 1.29+, Helm chart (oci://registry.promptglass.ai/charts/promptglass), PostgreSQL, ClickHouse, S3-compatible object storage, NATS, OpenTelemetry.
- Integrations: Splunk, Microsoft Sentinel, Elastic, QRadar, Chronicle, Slack, Teams, ServiceNow, Jira, REST API, MCP server (@promptglass/mcp).
- Pricing: free 30-day AI usage assessment; then per monitored employee per month, billed annually.

## Main pages
- [Home](https://promptglass.ai/)
- [Platform](https://promptglass.ai/platform)
- [How it works](https://promptglass.ai/how-it-works)
- [Deployment options](https://promptglass.ai/deployment)
- [Pricing](https://promptglass.ai/pricing)
- [Security and privacy](https://promptglass.ai/security)
- [AI cost calculator](https://promptglass.ai/ai-cost-calculator)
- [Contact](https://promptglass.ai/contact)

## Capabilities
- [AI usage monitoring](https://promptglass.ai/capabilities/ai-usage-monitoring): Who uses which AI tool, how often, for what, at what cost.
- [Prompt & response capture](https://promptglass.ai/capabilities/prompt-capture): Searchable record of prompts, responses and uploads, under your privacy rules.
- [AI data loss prevention](https://promptglass.ai/capabilities/ai-data-loss-prevention): Detect PII, PHI, secrets and code in prompts. Coach, redact or block inline.
- [Shadow AI discovery](https://promptglass.ai/capabilities/shadow-ai-discovery): Find unapproved AI apps, personal accounts, extensions and local models.
- [AI cost management](https://promptglass.ai/capabilities/ai-cost-management): Spend by team and model, idle seats, duplicate tools, right-sizing.
- [AI productivity & efficiency](https://promptglass.ai/capabilities/ai-productivity-analytics): Use cases, prompt quality, rework rates, and where training pays off.
- [AI policy enforcement](https://promptglass.ai/capabilities/ai-policy-enforcement): Allow, coach, redact, block or reroute, per app, group and data class.
- [AI gateway & proxy](https://promptglass.ai/capabilities/ai-gateway): OpenAI/Anthropic-compatible proxy with keys, budgets, routing and DLP.

## AI tools monitored
- [Monitor ChatGPT](https://promptglass.ai/ai-tools/chatgpt): Free, Plus, Team and Enterprise workspaces, web and desktop.
- [Monitor Claude](https://promptglass.ai/ai-tools/claude): Claude.ai Free, Pro, Team and Enterprise, plus Claude desktop and Claude Code.
- [Monitor Microsoft 365 Copilot](https://promptglass.ai/ai-tools/microsoft-365-copilot): Copilot in Word, Excel, Outlook, Teams and Copilot Chat.
- [Monitor GitHub Copilot](https://promptglass.ai/ai-tools/github-copilot): Copilot Business and Enterprise in VS Code, JetBrains and github.com.
- [Monitor Google Gemini](https://promptglass.ai/ai-tools/google-gemini): Gemini app, Gemini in Workspace, and the Gemini API.
- [Monitor Cursor](https://promptglass.ai/ai-tools/cursor): AI-first code editor with chat, agents and multi-model access.
- [Monitor Perplexity](https://promptglass.ai/ai-tools/perplexity): AI answer engine with web search, Spaces and file uploads.
- [Monitor DeepSeek](https://promptglass.ai/ai-tools/deepseek): Chat app and API for DeepSeek models.
- [Monitor Microsoft Copilot (consumer)](https://promptglass.ai/ai-tools/microsoft-copilot-chat): copilot.microsoft.com and the Copilot app, signed in with personal or work accounts.
- [Monitor Grok](https://promptglass.ai/ai-tools/grok): xAI assistant on the web, in apps and through the API.
- [Monitor Mistral Le Chat](https://promptglass.ai/ai-tools/mistral-le-chat): Le Chat assistant and La Plateforme API.
- [Monitor Notion AI](https://promptglass.ai/ai-tools/notion-ai): AI built into Notion pages, databases and Q&A.
- [Monitor OpenAI API](https://promptglass.ai/ai-tools/openai-api): Responses, Chat Completions, Assistants and embeddings APIs.
- [Monitor Anthropic API](https://promptglass.ai/ai-tools/anthropic-api): Messages API, tool use and batch for Claude models.
- [Monitor Azure OpenAI](https://promptglass.ai/ai-tools/azure-openai): OpenAI models hosted in your Azure subscription.
- [Monitor Amazon Bedrock](https://promptglass.ai/ai-tools/aws-bedrock): Foundation models from several providers in your AWS account.
- [Monitor Local LLMs (Ollama, LM Studio)](https://promptglass.ai/ai-tools/local-llms): Models running on laptops and workstations, outside network controls.

## Industries
- [Healthcare](https://promptglass.ai/industries/healthcare)
- [Financial services](https://promptglass.ai/industries/financial-services)
- [Legal](https://promptglass.ai/industries/legal)
- [Government & public sector](https://promptglass.ai/industries/government)
- [Software & technology](https://promptglass.ai/industries/technology)
- [Manufacturing](https://promptglass.ai/industries/manufacturing)

## Compliance frameworks
- [EU AI Act](https://promptglass.ai/compliance/eu-ai-act)
- [ISO/IEC 42001](https://promptglass.ai/compliance/iso-42001)
- [NIST AI Risk Management Framework](https://promptglass.ai/compliance/nist-ai-rmf)
- [GDPR](https://promptglass.ai/compliance/gdpr)
- [HIPAA](https://promptglass.ai/compliance/hipaa)
- [SOC 2](https://promptglass.ai/compliance/soc-2)
- [PCI DSS](https://promptglass.ai/compliance/pci-dss)

## Comparisons
- [PromptGlass vs Microsoft Purview](https://promptglass.ai/compare/microsoft-purview-alternative)
- [PromptGlass vs Netskope](https://promptglass.ai/compare/netskope-alternative)
- [PromptGlass vs Zscaler](https://promptglass.ai/compare/zscaler-alternative)
- [PromptGlass vs Nightfall AI](https://promptglass.ai/compare/nightfall-alternative)

## Glossary
- [AI gateway](https://promptglass.ai/learn/ai-gateway): An AI gateway (also called an LLM proxy or LLM gateway) is a proxy that sits between users or applications and AI model providers. It authenticates callers, records usage and cost, routes requests to the right model, and applies security controls such as data-leakage detection, without changing application code.
- [AI data loss prevention (AI DLP)](https://promptglass.ai/learn/ai-data-loss-prevention): AI data loss prevention (AI DLP) is a set of controls that inspect the content people and applications send to AI tools (prompts, pasted text and uploaded files) and stop sensitive data from leaving the organization, by warning the user, redacting the sensitive parts, blocking the request, or routing it to an approved model.
- [AI acceptable use policy](https://promptglass.ai/learn/ai-acceptable-use-policy): An AI acceptable use policy is an organizational policy that states which AI tools employees may use, for which purposes, with which categories of data, and under what conditions, along with how use is monitored and what happens when the policy is broken.
- [AI governance](https://promptglass.ai/learn/ai-governance): AI governance is the set of policies, roles, processes and tools an organization uses to make sure AI is used safely, legally, cost-effectively and in ways that create value. For most organizations, the largest and fastest-growing part is governing how employees use generative AI tools.
- [LLM token costs](https://promptglass.ai/learn/llm-token-costs): LLM token costs are the charges model providers bill for API use, calculated from the number of input tokens (the prompt and context) and output tokens (the response), each at a per-model rate. Tokens are word fragments; long contexts, agent loops and large models drive cost up.
- [Prompt injection](https://promptglass.ai/learn/prompt-injection): Prompt injection is an attack in which instructions are hidden in content that an AI model processes, such as a web page, email, document or tool result, so that the model follows the attacker’s instructions instead of its intended ones, for example by leaking data or taking unwanted actions.
- [AI literacy](https://promptglass.ai/learn/ai-literacy): AI literacy is the combination of skills, knowledge and understanding that lets people use AI systems effectively and responsibly: knowing what a tool can and cannot do, how to prompt it well, how to check its output, and which data must never be shared with it.

## Documentation
- [PromptGlass documentation](https://promptglass.ai/docs): Install, configure and operate PromptGlass, the AI usage monitoring platform for prompt capture, AI DLP, cost governance and adoption analytics.
- [Adoption analytics](https://promptglass.ai/docs/analytics/adoption): Measure AI adoption with PromptGlass. Track active users, teams, apps and use cases over time, and find the champions and laggards in your rollout.
- [Cost optimization](https://promptglass.ai/docs/analytics/cost-optimization): Track AI spend by team, model and app, find idle ChatGPT Enterprise, Copilot and Claude seats, and right-size models with PromptGlass recommendations.
- [Efficiency analytics](https://promptglass.ai/docs/analytics/efficiency): Understand how effectively employees use AI with prompt quality scores, re-prompt rates, use-case clustering and transparent time-saved estimates.
- [Shadow AI discovery](https://promptglass.ai/docs/analytics/shadow-ai): Discover unsanctioned AI apps, browser extensions, local models and AI features inside SaaS tools, and score each one's risk with the PromptGlass AI catalog.
- [Architecture](https://promptglass.ai/docs/architecture): How PromptGlass sensors, the AI gateway, detection workers, policy engine and analytics store fit together, and what data is stored where.
- [Anthropic (Claude)](https://promptglass.ai/docs/connectors/anthropic): Connect Claude Enterprise and Team workspaces and Anthropic API organizations to PromptGlass for usage, cost, seat utilization and compliance data.
- [AWS Bedrock](https://promptglass.ai/docs/connectors/aws-bedrock): Connect AWS Bedrock to PromptGlass using CloudTrail and model invocation logging for usage, cost and prompt-level detection across accounts.
- [Azure OpenAI](https://promptglass.ai/docs/connectors/azure-openai): Connect Azure OpenAI and Azure AI Foundry deployments to PromptGlass with diagnostic settings, Azure Monitor metrics and cost data, or route through the gateway.
- [GitHub Copilot](https://promptglass.ai/docs/connectors/github-copilot): Connect GitHub Copilot Business and Enterprise to PromptGlass to measure seat use, acceptance rates, language adoption and idle seats by team.
- [Google Gemini for Workspace](https://promptglass.ai/docs/connectors/google-gemini): Connect Google Workspace audit logs to PromptGlass to report Gemini usage in Gmail, Docs, Sheets, Meet and the Gemini app by user and organizational unit.
- [Microsoft 365 Copilot](https://promptglass.ai/docs/connectors/microsoft-365-copilot): Connect PromptGlass to Microsoft 365 Copilot through Microsoft Graph and the Purview audit log to report usage, seat utilization and sensitive interactions.
- [OpenAI (ChatGPT Enterprise and API)](https://promptglass.ai/docs/connectors/openai): Connect ChatGPT Enterprise and Edu workspaces via the Compliance API, and OpenAI API organizations via the usage and costs APIs, to PromptGlass.
- [High availability](https://promptglass.ai/docs/deployment/high-availability): Run PromptGlass with no single point of failure using multiple gateway replicas, replicated ClickHouse, HA PostgreSQL and per-policy fail-open or fail-closed.
- [Deploy on Kubernetes](https://promptglass.ai/docs/deployment/kubernetes): Self-host PromptGlass on Kubernetes 1.29+ with the official Helm chart, including values.yaml, ingress, autoscaling and external PostgreSQL or ClickHouse.
- [On-premises and air-gapped](https://promptglass.ai/docs/deployment/on-premises-air-gapped): Install PromptGlass with no internet connectivity using an offline bundle, a private registry, local AI catalog updates and an on-premises HSM.
- [SaaS deployment](https://promptglass.ai/docs/deployment/saas): Run PromptGlass as a managed SaaS service in the US, EU or Canada region, with private connectivity options and customer-managed encryption keys.
- [Custom detectors](https://promptglass.ai/docs/detection/custom-detectors): Define your own PromptGlass detectors with regex, validators, dictionaries, keyword proximity and ML classifiers to catch organization-specific data.
- [Built-in detectors](https://promptglass.ai/docs/detection/detectors): Reference for PromptGlass built-in AI DLP detectors covering PII, PHI, PCI, secrets and credentials, source code and confidential business data.
- [Exact data match](https://promptglass.ai/docs/detection/exact-data-match): Use PromptGlass exact data match (EDM) to detect your real customer records and proprietary code in prompts using salted hashes, never plaintext.
- [Alerts and ticketing](https://promptglass.ai/docs/integrations/alerts-and-ticketing): Send PromptGlass alerts to Slack, Microsoft Teams, email, PagerDuty and webhooks, and open ServiceNow or Jira tickets for AI incidents and seat reclaims.
- [SSO and SCIM](https://promptglass.ai/docs/integrations/identity-sso-scim): Connect PromptGlass to Entra ID, Okta or Google Workspace for single sign-on and SCIM provisioning to attribute AI usage to people, teams and cost centers.
- [MCP server](https://promptglass.ai/docs/integrations/mcp-server): Install the PromptGlass MCP server to ask Claude, Copilot or Cursor questions about AI adoption, spend and risk in your organization. Read-only by default.
- [REST API](https://promptglass.ai/docs/integrations/rest-api): Use the PromptGlass REST API to query AI usage events, adoption and cost metrics, manage policies and detectors, and export data to your lakehouse.
- [SIEM integration](https://promptglass.ai/docs/integrations/siem): Stream PromptGlass AI usage events and detections to Splunk, Microsoft Sentinel, Elastic, IBM QRadar or Google Chronicle over HTTP, syslog or CEF.
- [Policy examples](https://promptglass.ai/docs/policies/examples): Ready-to-adapt PromptGlass policy examples for secrets, source code, customer data, HIPAA PHI, shadow AI coaching, budgets and routing to approved models.
- [Policies overview](https://promptglass.ai/docs/policies/overview): How PromptGlass policies work, including rule order, scopes, actions from allow to block and route, coaching messages, testing and GitOps workflows.
- [Policy reference](https://promptglass.ai/docs/policies/policy-reference): Complete YAML schema for PromptGlass policies including match conditions, actions, storage modes, budgets, messages, exceptions and error handling.
- [Data residency and retention](https://promptglass.ai/docs/privacy/data-residency-and-retention): Control where PromptGlass stores AI usage data, how long it keeps prompts and metadata, and how content is encrypted, exported and deleted.
- [Employee privacy](https://promptglass.ai/docs/privacy/employee-privacy): How PromptGlass protects employees with pseudonymization by default, four-eyes re-identification, works-council mode, k-anonymity and transparency notices.
- [Quickstart](https://promptglass.ai/docs/quickstart): Go from a new PromptGlass SaaS tenant to a live AI usage dashboard in about 15 minutes with SSO, the browser extension and one connector.
- [Changelog](https://promptglass.ai/docs/reference/changelog): Release notes for PromptGlass covering the gateway, browser extension, endpoint agent, connectors, detectors, policies, analytics and the Helm chart.
- [CLI reference](https://promptglass.ai/docs/reference/cli): Reference for the promptglass command-line tool, covering login, policies, detectors, connectors, gateway status, keys, exports, catalog and air-gap commands.
- [FAQ](https://promptglass.ai/docs/reference/faq): Answers to common questions about PromptGlass deployment, prompt capture, TLS inspection, employee privacy, performance, supported AI tools and pricing.
- [Helm values reference](https://promptglass.ai/docs/reference/helm-values): Reference for the PromptGlass Helm chart values covering global settings, gateway, detectors, auth, encryption, data stores, ingress and telemetry.
- [Browser extension](https://promptglass.ai/docs/sensors/browser-extension): Deploy the PromptGlass browser extension to Chrome, Edge, Firefox and Safari with Intune, Jamf, Google Admin or Group Policy to capture web AI usage.
- [Endpoint agent](https://promptglass.ai/docs/sensors/endpoint-agent): Use the PromptGlass endpoint agent on Windows, macOS and Linux to discover desktop AI apps, IDE coding assistants and local models like Ollama.
- [Gateway API mode](https://promptglass.ai/docs/sensors/gateway-api): Point OpenAI and Anthropic SDKs at the PromptGlass Gateway for attribution, AI DLP, budgets, model routing and per-team virtual API keys.
- [Gateway forward proxy](https://promptglass.ai/docs/sensors/gateway-forward-proxy): Route browser and desktop AI traffic through the PromptGlass Gateway with a PAC file or HTTPS_PROXY, inspecting TLS only for domains in the AI catalog.
- [Chaining behind your SWG](https://promptglass.ai/docs/sensors/swg-chaining): Forward only AI traffic from Zscaler, Netskope or Palo Alto Prisma Access to the PromptGlass Gateway, keeping your existing secure web gateway in place.

## Blog
- [AI Gateway vs Browser Extension vs CASB: Which Do You Need?](https://promptglass.ai/blog/ai-gateway-vs-browser-extension-vs-casb): Compare AI gateways, browser extensions, CASB/SWG and vendor APIs for monitoring and controlling employee AI use, covering coverage, inline control and privacy.
- [EU AI Act AI Literacy: What It Means for Employee AI Use](https://promptglass.ai/blog/eu-ai-act-ai-literacy-employee-ai-usage): What the EU AI Act's AI literacy duty and deployer obligations mean for everyday employee use of AI tools, and how usage data helps you show progress.
- [How to Cut AI Spend: Find Unused Copilot and ChatGPT Seats](https://promptglass.ai/blog/reduce-ai-spend-unused-copilot-chatgpt-seats): A step-by-step method to find idle Microsoft 365 Copilot, ChatGPT Enterprise, Claude and GitHub Copilot seats, remove duplicates, and right-size API models.
- [AI Data Leakage Prevention: A Practical Guide to AI DLP](https://promptglass.ai/blog/ai-data-leakage-prevention-guide): How sensitive data leaks into AI tools, why traditional DLP misses it, and how to detect, redact and block it inline without killing productivity.
- [How to Monitor ChatGPT Usage in Your Company (2026 Guide)](https://promptglass.ai/blog/how-to-monitor-chatgpt-usage-in-your-company): A practical guide to monitoring ChatGPT use at work, covering Enterprise workspaces, personal accounts and the API, plus what to capture and how to stay privacy-friendly.
- [What Is Shadow AI? Definition, Risks and How to Find It](https://promptglass.ai/blog/what-is-shadow-ai): Shadow AI is any AI tool used for work without IT approval. Learn the common forms, the real risks, and a practical way to discover and manage it.