Endpoint Authority
Find and control AI agents on every laptop
Endpoint Authority is a sensor on employee laptops that finds the AI agents and MCP servers people run and stops risky actions before they happen: four pillars built as one system.
At a glance
What's in Endpoint Authority
- Why it's differentDevice, person, agent and server, checked on the laptopOne on-device decision
- Shadow AI detectionEvery agent on each laptop, including ones people set up themselvesInstalled agents · Where they send data
- MCP server detectionLocal MCP servers off your approved list, flaggedLocal servers · Outside connections
- Agent identityEach agent tied to a verified person and groupVerified person · Offboarding
- On-device controlRisky actions stopped on the laptop, before they runOn-device checks · Agent Trail
On-device decisions
Decided on the laptop, with a name behind it
Some agent actions never pass through a gateway. Endpoint Authority sees them where they start, on the laptop, and knows the verified person behind every agent.
- Finds the agents and local MCP servers people set up on their own
- Matches each one to the person signed in through your identity provider
- Checks your rules before the action runs, and logs every decision
- Sensor active
Device
MBP-0412
macOS laptop
- Verified
Person
Platform engineer
Group eng-platform, from Okta
- Shadow AI
Agent
Cursor
Installed by the user
- Unapproved MCP
Local MCP server
local-postgresLocal, used by Cursor, not on your approved list
- Rule matched
Check
Only approved MCP servers
On-device check, before it runs
- Blocked
Decision
$ cursor connect mcp://local-postgres
Stopped on the laptop, before it runs
Pillar 1 · Shadow AI detection
Find every agent running on each laptop
The sensor shows which agents are installed and running on each laptop, including the ones people set up on their own.
- Claude Code, Cursor, Copilot, Codex, Gemini CLI and desktop AI apps
- Where each agent sends data, so a new destination stands out
Pillar 2 · MCP server detection
Find shadow MCP servers and see where they connect
People add MCP servers to their agents on their own. Endpoint Authority finds them on the laptop and shows where they connect.
- Local servers flagged when they aren't on your approved list
- Connections to MCP servers outside your company, with the agent that opened them
Pillar 3 · Agent identity
Every agent on a laptop is tied to a verified person
Each device and agent is matched to the person signed in through your identity provider. When something goes wrong, you start from a name and a group, not a hostname.
- Rules can follow your org chart, using the groups you already have
- When someone leaves, their agents lose access with them
- Every agent, device and action in one trail
Pillar 4 · On-device control
Stop the risky action on the laptop, before it happens
Endpoint Authority applies your rules right on the device, so a local agent like Cursor can't connect to an unapproved MCP server or run something it shouldn't, even when nothing touches a gateway.
- Each blocked action shows up in Agent Trail with the person and the rule
- The same rules for every agent across Claude Code, Cursor, Copilot and Codex
Part of Agent Authority
One authority for every path an agent takes
Endpoint Authority is one of three enforcement paths in Agent Authority, alongside MCP Authority and LLM Authority.
Where AI runs
Agent Authority
One policy on every path
What it reaches
Employees and desktop agents
Endpoint Authority
Discover AI apps, block unsanctioned tools
Local files and shadow AI
Agents calling tools over MCP
MCP Authority
Authorize every tool call, scoped to the task
Enterprise apps and data
Prompts from users and agents
LLM Authority
Redact sensitive data, stop prompt injection
Model providers
Endpoint Authority FAQ
What is Endpoint Authority?
Endpoint Authority is the part of Agent Authority that works on employee laptops. A sensor finds the AI agents and local MCP servers people run, ties each agent to the verified person signed in through your identity provider, and applies your rules on the device, before an action runs.
Which AI agents can Endpoint Authority see?
Agents such as Claude Code, Claude Desktop, Cursor, GitHub Copilot, Codex and Gemini CLI, plus desktop AI apps and local MCP servers. It shows the ones IT installed and the ones people set up on their own, and where each agent sends data, so a new destination stands out.
How does Endpoint Authority find shadow MCP servers?
Local MCP servers are flagged when they aren't on your approved list. Connections from a laptop to MCP servers outside your company show up too, with the agent that opened them.
Why enforce policy on the device instead of at a gateway?
Some agent actions never pass through a gateway, such as a local agent connecting to a local MCP server or running a command. Endpoint Authority checks them on the device before they run, with the same rules for every agent across Claude Code, Cursor, Copilot and Codex.
What happens to someone's agents when they leave?
Each agent is tied to the person signed in through your identity provider, so when someone is offboarded, for example in Okta, their agents lose access with them.
How do we roll out Endpoint Authority?
With the tools you already manage laptops with, such as Jamf and Intune. There's no new browser and no proxy change, and once your identity provider is connected, each laptop maps to a person.
Where are blocked actions recorded?
In Agent Trail, with the person, the agent and the rule. Agent Trail keeps every agent, device and action in one trail.
See every AI agent on every device
14-day free trial. Your own Agent Authority tenant. Roll it out with the tools you already manage laptops with: read the setup docs.
Start free trial