Agent Identity
Know every agent, and who owns it.
- Agents found as they are used, not by survey
- An owner and a team on each one
- Every action attributed to a person
NewOn-device data classification
Scale AI with controlacross your company.
The AI governance control plane that finds every AI agent in your company, gives each one an owner and limits, and shows what it costs. Live in your browsers within an hour, and nothing anyone types leaves the device.
Works with
Every agent, and the stack it runs on
ChatGPTAI appClaudeAI appGeminiAI appCopilotAI appClaude CodeDev toolCursorDev toolCodexDev toolCustom GPTsAgents
ChatGPTAI appClaudeAI appGeminiAI appCopilotAI appClaude CodeDev toolCursorDev toolCodexDev toolCustom GPTsAgents
ChatGPTAI appClaudeAI appGeminiAI appCopilotAI appClaude CodeDev toolCursorDev toolCodexDev toolCustom GPTsAgents
CrewAIFrameworkLangGraphFrameworkGoogle ADKFrameworkAgnoFrameworkLiteLLMGatewayOpenTelemetryTelemetryGoogle AdminMDMIntuneMDMJamfMDM
CrewAIFrameworkLangGraphFrameworkGoogle ADKFrameworkAgnoFrameworkLiteLLMGatewayOpenTelemetryTelemetryGoogle AdminMDMIntuneMDMJamfMDM
CrewAIFrameworkLangGraphFrameworkGoogle ADKFrameworkAgnoFrameworkLiteLLMGatewayOpenTelemetryTelemetryGoogle AdminMDMIntuneMDMJamfMDM
The three pieces missing
from your AI rollout.
Agents are already in every team. Ownership, limits and findings are not.
Agent Governance
Decide what each agent may do.
- A limit per agent: tools, data, autonomy
- Set once by IT, applied to every team
- One-sentence alert when it is crossed
Agent Security
Turn every violation into a finding.
- Data leaks, crossed limits, failed scans
- One inbox, mapped to your frameworks
- Detected without content leaving the device
The first scan
One team, one week.
Every agent on the table.
Install Lustena on one team. In a week it has found every agent that team uses, ranked what to govern first and shown where an official agent would pay off.
- 01Every agent, with an ownerApps, custom GPTs, projects, connectors and code agents, each attributed to the people who use it.
- 02What to govern firstA ranked list by data exposure, what each agent can reach and how much it is used, with the next action on each.
- 03Where to deploy nextThe work each team already brings to AI on its own: the demand an official agent should meet.
- 04What it costsPaid seats, idle seats and token spend per agent, with an owner for every line.
Lustena extends your stack
Lustena connects to the browsers, device management, frameworks and gateways you already run, so agents come under the same controls as your people.
Browsers & devices
ChromeEdgeDesktop appsGoogle AdminIntuneJamf+ any Chromium browser
Developer tools
Claude CodeCursorCodex+ any MCP client
Agent frameworks
CrewAILangGraphGoogle ADKAgno+ any OpenTelemetry source
AI gateways
LiteLLM+ more on request
Alerts & audit
WebhooksEvidence package+ any webhook receiver: Slack, SIEM
I thought we had five agents. We have seven. Two of them can write to our CRM, and nobody owns them.
One golden path
for every agent.
Whether an agent acts for an employee in a browser tab or runs on its own inside your stack, Lustena brings it under one identity layer, one set of limits and one audit trail.
Read the guide: what is an AI governance control plane?Employee agents
Made by anyone, without code, inside ChatGPT, Claude, Gemini or Copilot: a GPT for proposals, a project with the finance sheet, a connector to Gmail. IT never sees them being made.
Custom GPTsClaude projectsConnectorsCopilot Studio bots
Seen by the browser extension
01Where the agent acts · browser
Client follow-up
Paste the client list from the spreadsheet and draft the follow-up email for each.
Reply
02Lustena, in the path
Lustena· Agent session live
ana@acme.com · GPT · Proposals
- Agent Identitydeclared · owner anaok
- Agent Security1 paste · 3 × card number on deviceflagged
- Agent Governancelimit v3 · data class: personalflagged
03Systems reached
- Gmailconnector
- Driveconnector
- HubSpotconnector
Your people are already AI-native.
Your governance should be too.
Lustena sits where agents already act and applies identity, limits, security and observability to every session.
Lustena· Agents
Agents
- GPT · Proposalsana@acme.comowned
- Claude · Finance projectjoao@acme.comowned
- Copilot Studio · HR FAQno owner yetobserved
Every team ships agents.
IT keeps control.
Marketing, sales and finance put agents into production without a ticket queue and without hiring AI specialists to govern them. IT gets the answers that let it say yes: what is running, what it can reach and what it costs.
See it on your agents- <0min
- To your first agent inventory
- 0%
- Of sessions checked against a limit
- 0
- Place to see AI adoption, licenses and cost
Publish the extension policy in Google Admin, Intune or Jamf. Agents appear as they are used, attributed to the person using them, with no survey and no change to the agents.
Every session seen by the browser extension, the SDK or the MCP server is evaluated against its agent's limit, or the workspace default, from day one, with no setup per project.
Sessions per app, team and agent, with model and volume per agent. Cut unused licenses and give every token an owner.
Privacy and compliance by design
Compliant from the
first agent, not the
last audit
Teams that build and use agents connected to Lustena are in compliance while they work. Every agent has an owner and a limit before it ships, content stays on the device by default, and the evidence an auditor asks for is written as a by-product of every session, not reconstructed in a panic the week before.
After the fact
- Agents ship first; someone builds the inventory from a survey months later
- Policy is a PDF nobody can check an agent against
- Employee prompts are logged in full, creating a new privacy risk to manage
- Evidence is screenshots and email threads gathered for the audit
By design, with Lustena
- An agent appears with an owner the first time it is used or deployed
- A versioned limit is evaluated on every session, from day one
- Classification runs on the device; only content-free findings leave it
- The evidence package is current every day, mapped to the framework you name
Ready for certification
Lustena prepares the evidence each framework asks for. Independent assessors certify; Lustena makes the assessment faster.
Regulation
GDPR & LGPD
Content-free findings by default, a written lawful basis before any capture, retention limits and audited reads.
Trust Center →RegulationEU AI Act
Support for risk classification, an inventory with an owner for every agent, logging and human oversight evidence for each agent you run.
Read the guide →CertificationISO/IEC 42001
A governed agent inventory, declared limits and a decision history for the controls of your AI management system.
Read the guide →FrameworkNIST AI RMF
Govern, map, measure and manage, with fleet-wide visibility, per-agent evaluation and runtime alerts.
Read the guide →CertificationAIUC-1
Evidence across the six domains, from data and privacy to accountability, organized for an independent AIUC-1 assessment.
Read the guide →FrameworkOWASP LLM Top 10
Red-team scans with ten agent-native checks, each mapped to the risk it covers, attached to the agent.
See the scans →Compliance at Lustena
The control plane holds itself to the rule it enforces: by default, prompt text, file contents and file names are never stored or sent. Named standards are references; Lustena does not issue certifications and does not claim ones it does not hold.
Content-free by constructionPrivacy by designAudited readsRetention limits
Visit the trust centerQuestions
What is an AI governance control plane?
A central layer that gives you identity, policy and audit over every AI agent your company runs, whether engineers built it or an employee assembled it inside ChatGPT. Lustena's control plane starts in the browser, where most of that use already happens, and extends to code-built agents through a Python SDK, an MCP server and a scanner.
What AI agents does Lustena support?
In the browser: ChatGPT, Claude, Gemini and Copilot. On the desktop, on the Enterprise plan: the ChatGPT, Claude and Copilot apps. In code: CrewAI, LangGraph, Google ADK, Agno and anything that emits OpenTelemetry. In developer tools: Claude Code, Cursor and Codex through the MCP server. Each agent appears in one inventory under one policy layer.
How does Lustena handle shadow AI?
The extension sees AI use as it happens on managed browsers, without content leaving the device. Each app, custom GPT or connector becomes an agent in the inventory with an owner. Developer tools register themselves when the MCP server is installed, and on the Enterprise plan the desktop collector covers the ChatGPT, Claude and Copilot desktop apps. What remains outside coverage is shown as a number, not hidden. Shadow AI is also a demand signal: the work people already bring to AI on their own shows where an official agent should go next.
How are AI policies defined?
As limits per agent across six dimensions: tools, MCP servers, hosts, data classes, autonomy level and schedule. Limits are versioned, and the history of what changed and who changed it is kept.
How are policies enforced?
Every session is evaluated against the agent's limits. An externally run agent that crosses a limit is flagged with one sentence saying what crossed it, and alerts go out through webhooks. Agents running on the Lustena SDK runtime can be stopped before the call is made, and in the browser the Enterprise plan can warn, block or redact before a prompt is sent.
Does Lustena replace my identity provider?
No. Devices enroll with an enrollment token and carry the employee's corporate email, and agents get their own identity on top of the people who use them. Platform access uses passwordless login, with SSO through your identity provider on the Enterprise plan.
How does Lustena handle security and compliance?
Classification runs on the device and only content-free findings leave it. Content capture is off by default and, when turned on, requires a written lawful basis, is redacted on the device, encrypted at rest and audited on every read. Evidence packages map findings to GDPR, the EU AI Act, ISO/IEC 42001, NIST AI RMF, AIUC-1 and LGPD. Lustena prepares the evidence; independent assessors certify.
How long does it take to deploy?
Most teams see their first agents within an hour: publish the extension policy in Google Admin, Intune or Jamf, and devices enroll on their own. Nothing changes in the agents themselves.
Do I need to install anything inside my agents?
No. The browser extension needs nothing from the agent. For code-built agents, the Python SDK instruments your framework in a few lines, and the scanner only needs an endpoint you are authorized to test.
Can Lustena be self-hosted?
Yes. Lustena runs as a managed service on Google Cloud, and on the Enterprise plan the control plane can be deployed in your own cloud or on premises.