llms.txt
Compare with other sites' llms.txt files →# Privent
> Privent is the AI data control layer inside your agent runtime.
It embeds in your n8n workflows, replaces sensitive values with reversible tokens before they reach any external model, and produces audit evidence your customers can verify. Self-host, on-prem, or air-gapped. n8n is available today via the official node, n8n-nodes-privent. The @priventai/mcp-server package gives any MCP client the tokenize, risk, and audit tools, and the @priventai/mcp-gateway package tokenizes MCP tool calls in front of any stdio server you run. Links ending in .md are plain-markdown twins of the HTML page.
## When to use Privent
- An n8n or MCP workflow sends customer, patient, or financial data to an external model with no human in the loop.
- A client or auditor needs evidence of what left the boundary, per session and per step.
- Data must stay inside a self-hosted, on-prem, or air-gapped deployment.
## Product
- [Privent - The AI Data Control Layer for Agentic AI](https://privent.ai/): Privent is the AI data control layer for agentic AI: reversible tokens inside your agent runtime, before any external model sees the data, with a signed audit report.
- [Agent Security for n8n](https://privent.ai/agent-security): The Privent n8n node replaces sensitive values with typed, reversible tokens inside the workflow, before any external model sees them, and restores them at the trusted sink.
- [The n8n In-Engine Interceptor](https://privent.ai/interceptor): The Privent interceptor is a build of the n8n image that transforms node egress and agent tool-call arguments inside the engine: one image swap, no workflow changes.
- [Security & Data Handling](https://privent.ai/security): Privent never stores raw prompt content, only metadata: tokens live in a vault you can self-host, and every claim on this page is stated the same way inside the product.
- [Pricing](https://privent.ai/pricing): Privent pricing is public: Pro from $19 and Teams from $179 per month, reversible tokenization in your n8n workflows on every tier, and every plan can run on-prem.
- [Privent for Automation Agencies](https://privent.ai/for-agencies): Privent lets an agency build n8n workflows on client data and hand the client a signed, attestable evidence report under the agency's own brand, mapped to 8 frameworks.
## Integrations
- [Integrations](https://privent.ai/integrations): 3 of 6 Privent surfaces install today (n8n in-engine interceptor, n8n and MCP Server); the other rows carry the status the product's own constants report, never a hand-written one.
- [LangGraph Integration (Coming Soon)](https://privent.ai/integrations/langgraph.md): PriventSecurityNode for LangGraph is coming soon: full graph state access, 6-signal ACARS risk scoring, and adaptive APE transformation. Available today for n8n.
- [CrewAI Integration (Coming Soon)](https://privent.ai/integrations/crewai.md): Privent middleware for CrewAI is coming soon: agent context, task memory, and inter-task payload interception. Available today for n8n.
- [n8n Integration](https://privent.ai/integrations/n8n.md): n8n-nodes-privent adds tokenize, risk-score and detokenize nodes to any n8n workflow, so sensitive values become reversible tokens before they reach an LLM node.
- [MCP Server Integration](https://privent.ai/integrations/mcp-servers.md): The @priventai/mcp-server package gives any MCP client Privent's tools, and @priventai/mcp-gateway tokenizes MCP tool calls in front of any stdio server you run.
- [Custom SDK (Coming Soon)](https://privent.ai/integrations/custom-sdk.md): The Privent SDK is coming soon: ACARS risk scoring and APE transformation as a library for any Python or JavaScript agent. Available today for n8n.
## Use cases
- [Use Cases](https://privent.ai/use-cases): Privent use cases show where n8n agent workflows meet sensitive data and how reversible tokens keep the workflow running while the data stays inside the boundary.
- [Patient Data Leakage in Agent Workflows](https://privent.ai/use-cases/agentic-ai-data-leakage.md): n8n intake, triage and scheduling workflows send patient records to external models with no one reviewing; Privent tokenizes PHI inside the workflow and restores it at the sink.
- [Credential & API Key Exposure](https://privent.ai/use-cases/credential-api-key-exposure.md): Agents wired to EHR and FHIR APIs surface keys, passwords and database credentials in their prompts; Privent catches them inside the workflow before they leave your network.
- [HIPAA Audit Trails for AI](https://privent.ai/use-cases/ai-compliance-audit-trails.md): Every Privent detection event is logged with its risk signals, category and decision, never the raw PHI, as exportable evidence for HIPAA reviews, GDPR and the EU AI Act.
- [Cross-Agent Memory Leakage](https://privent.ai/use-cases/cross-agent-memory-leakage.md): Patient context accumulates across agent steps and crosses between agents as shared memory; Privent controls what may cross at every boundary and tokenizes what may not.
- [Control Financial Data in n8n AI Workflows](https://privent.ai/use-cases/financial-services-ai-workflows.md): n8n onboarding and collections workflows send customer and payment data to AI models; Privent tokenizes sensitive values before any external model and restores them at the sink.
- [Protect PHI in n8n AI Workflows](https://privent.ai/use-cases/healthcare-ai-workflows.md): Patient-facing n8n workflows send PHI to external models; Privent tokenizes patient data inside the workflow and restores it only at trusted systems, so plaintext PHI never leaves.
- [Prove Safe Client Data Handling in n8n](https://privent.ai/use-cases/agency-client-workflows.md): Automation agencies build n8n workflows on client data; Privent tokenizes the values in the run and produces an attested, shareable evidence report under the agency's own brand.
- [Self-Hosted AI Workflow Security for n8n](https://privent.ai/use-cases/regulated-self-host-ai-workflows.md): For regulated teams whose data cannot leave the boundary: the Privent stack self-hosted, raw text never stored, keys in your custody, and audit evidence for your compliance team.
## Docs
- [Docs](https://privent.ai/docs.md): Privent documentation: install the n8n community nodes, configure org policy, run document tokenization jobs, and understand ACARS risk scoring. Self-host guides included.
- [Extension Setup](https://privent.ai/docs/extension-setup.md): Deploy the Privent browser extension through MDM managed policy or an activation token, connect your org API key, and verify capture across AI tools.
- [n8n Setup](https://privent.ai/docs/n8n-setup.md): Add the Privent community nodes to n8n Cloud or self-hosted, connect your org API key, and tokenize sensitive fields in your first workflow in minutes.
## Compare
- [Privent vs n8n Guardrails](https://privent.ai/vs/n8n-guardrails): n8n's free Guardrails node sanitizes text at the step it sits on and keeps no mapping back; Privent adds typed, reversible tokens across the whole pipeline, plus audit evidence.
## Blog
- [Blog](https://privent.ai/blog): Research and engineering notes on agentic AI security, prompt injection, runtime data control inside agent pipelines, and compliance from the Privent team.
- [How Agencies Prove Safe Data Handling to Clients](https://privent.ai/blog/how-agencies-prove-safe-data-handling.md): Learn how an automation agency answers a client security review with evidence: an anonymous scan, in-workflow tokenization, and a certified, revocable report.
- [n8n Guardrails: What It Does and Does Not Do](https://privent.ai/blog/n8n-guardrails-what-it-does-and-doesnt-do.md): Learn what the free n8n Guardrails node covers, the three gaps that matter for agent workflows, and how to tell when the node alone is enough.
- [Privent 2.0: Runtime Data Control for AI Agents](https://privent.ai/blog/privent-2-0-three-surfaces-one-vault.md): See what ships in Privent 2.0: an in-engine interceptor for n8n, an MCP gateway, and document tokenization, sharing one reversible token vault.
- [OpenAI Hugging Face Agent Incident, Explained](https://privent.ai/blog/openai-hugging-face-agent-security-incident-2026.md): Read what actually happened when an AI agent broke out of an OpenAI evaluation and reached Hugging Face production, and the four controls that stop it.
- [Agentic AI Security Incidents of June 2026](https://privent.ai/blog/agentic-ai-security-june-2026-enterprise-breach-warning.md): Get the June 2026 agentic AI incident rundown: SymJack, TrustFall, and Forrester's breach warning, plus the controls to put in place before agents go live.
- [Meta AI Agent Data Leak: What Happened](https://privent.ai/blog/meta-ai-agent-data-leak-pipeline-security-2026.md): Learn how Meta's internal AI agent exposed sensitive data with no attacker involved, why the same gap exists in agent pipelines, and how to close it.
- [Claude Self-Hosted Sandboxes and MCP Tunnels](https://privent.ai/blog/anthropic-self-hosted-sandboxes-mcp-tunnels-enterprise-ai-agents.md): See how Anthropic's self-hosted sandboxes and MCP tunnels work, and what they change for enterprise agent deployments blocked on data residency.
- [Patient Data in ChatGPT: The HIPAA Blind Spot](https://privent.ai/blog/healthcare-staff-chatgpt-patient-data-hipaa.md): Understand how clinical staff expose patient data through public AI tools, what HIPAA expects from you, and how to build a usage baseline before agents arrive.
- [AI Agent Memory Poisoning: OWASP ASI06 Explained](https://privent.ai/blog/memory-poisoning-crisis-enterprise-ai.md): Learn how memory poisoning plants instructions that steer agent decisions weeks later, why DLP misses it, and three defenses to deploy before scaling agents.
- [DLP for AI Agents: Why Current Controls Miss Them](https://privent.ai/blog/dlp-for-agentic-ai-pipelines.md): See why DLP built for humans cannot see agent pipelines, review the LangChain and LangGraph CVEs that prove it, and get a three-layer framework that can.
## Optional
- [About](https://privent.ai/about): Privent is three co-founders building the AI data control layer inside the agent runtime, so agents can move data at machine speed without leaking it.
- [Contact](https://privent.ai/contact): Talk to the Privent team about the AI data control layer inside your agent runtime: n8n deployment, fully on-prem installs, and shareable compliance evidence.
- [Privacy Policy](https://privent.ai/privacy): How Privent handles detection data: raw prompt content is never stored, transformation happens in-engine, and security teams keep full policy control.
- [Terms of Service](https://privent.ai/terms): Read the Terms of Service for using Privent, the agentic AI security platform that protects enterprise data across autonomous pipelines and AI tool usage.