# endlayer

> Every AI tool, agent, and model call across your company: surfaced, monitored, and governed from one plane.

endlayer is a product of Holistic AI. Contact: hello@endlayer.com.

## What it is

endlayer is an AI control plane. A local, unprivileged agent on each device reads what is already on the disk — installed AI apps, coding agents, MCP servers, local models, browser AI extensions, accounts and provider keys — and reports 21 signals every 900 seconds. There is no proxy in the path, no gateway to route through, and no per-tool connector.

The three layers, in order:

- **Surface** — find every AI tool, agent, model and key on the machine, without being told they exist.
- **Monitor** — what those tools reach, whose account they run on, which repositories they see, and what they cost.
- **Govern** — grade every signal against an immutable policy version, with an owner and a state for what fails, and apply the rules on the device: real-time guardrails in the agent warn, redact or block as it happens.

## What ships today

The device layer, backend and console ship today. Cloud connectors and SDK ingestion are planned.

The agent a company installs on each device monitors, detects and blocks. Every device is graded against a versioned policy, a failure gets an owner and a state, and the evidence is kept and exported. Real-time guardrails inside the agent act automatically on the device: prompts are scored before they run and secrets and personal data are warned on, redacted or blocked on the way out, with the matched text never leaving the machine. Removing an installed app is the one action that waits for a person: the Blocked apps rule raises a request an administrator approves before anything is uninstalled. Every action, automatic or approved, is recorded. Enforcement is not a roadmap item: every layer endlayer covers today is enforced, on the device, as it happens. Detection needs the agent; endlayer does not watch a device it is not installed on. Cloud connectors and SDK ingestion are the planned layers, and the same rules and the same enforcement apply to them the day they ship.

## Site structure

- [Home](https://www.endlayer.ai): what the product does, the console, the agent, and how it deploys
- [Platform](https://www.endlayer.ai/platform): how it works in depth — the three modules (#modules), the signal architecture (#architecture), the agent tiers (#agent), what it reads (#signals) and will not read (#boundary), the graph (#graph), policy grading (#governance), the record it keeps (#evidence), the console (#console), deployment (#deploy), self-hosting (#host), and what ships today (#today)
- [Surfaces](https://www.endlayer.ai/solutions/surface): the device layer. One row per device: the AI tools, agents, models and keys on each one
- [Monitoring](https://www.endlayer.ai/solutions/monitor): the fleet view. One row per AI tool across every device: who has it, what it reaches, whether it runs unattended, what it costs
- [Governance](https://www.endlayer.ai/solutions/govern): the rules. Nineteen versioned rules the fleet is graded against, real-time guardrails in the agent that warn, redact or block, approval-gated removal of installed apps, and an exportable record
- [Agents](https://www.endlayer.ai/agents): one agent, three tiers from 3 MB to 300 MB — Sensor, Decoder and Guardian — what each does and where it runs
- [The 21 signals](https://www.endlayer.ai/signals): every signal the agent reads, by group, with what each one reads and its schema key
- [Blog](https://www.endlayer.ai/blog): what is actually running on company laptops — unapproved AI tools, what coding agents may do on their own, keys in dotfiles, and what it all costs. Every post starts from a signal the agent reads. Full text of every post: https://www.endlayer.ai/llms-full.txt
- [Book a demo](https://www.endlayer.ai/demo): thirty minutes on a live console on 1,000 demo devices, and what the half hour covers
- [About](https://www.endlayer.ai/about): why endlayer exists, the problem it started from, and its place inside Holistic AI
- [Trust & security](https://www.endlayer.ai/trust): the controls behind the platform — ISO/IEC 27001:2022 certified ISMS, SOC 2 Type 2 report, UK GDPR, the 32 controls with their Annex A references, subprocessors, and the documents available under mutual NDA
- [Contact](https://www.endlayer.ai/contact): how to reach us
- [Topics](https://www.endlayer.ai/topics), [Authors](https://www.endlayer.ai/authors) and [Search](https://www.endlayer.ai/search): the blog's indexes
- [Sitemap](https://www.endlayer.ai/sitemap): every page, for people; [sitemap.xml](https://www.endlayer.ai/sitemap.xml) is the same list for search engines
- [Privacy Policy](https://www.endlayer.ai/privacy)
- [Terms of Service](https://www.endlayer.ai/terms)
- [Disclaimer](https://www.endlayer.ai/disclaimer)
- [Cookie Policy](https://www.endlayer.ai/cookies)

## The 21 signals

Read from files on the disk every fifteen minutes, by an unprivileged agent that never executes the tools it finds. Grouped as the product groups them:

- **Tools and agents** (`ai_tooling`, Tools & agents): Which AI tools, CLIs and coding agents are installed
- **Autonomy** (`ai_autonomy`, Tools & agents): Whether an agent can act on its own, and how far it is pre-approved
- **AI web usage** (`ai_web_usage`, Apps & web): Visits to known AI domains, by category and count
- **Installed software** (`installed_software`, Apps & web): The wider application inventory on the device
- **MCP servers** (`mcp_servers`, MCP & access): Declared servers, their transport, host and capability
- **Credentials** (`ai_credentials`, MCP & access): That a provider key exists, its variable name, and whether git ignores it
- **Accounts** (`ai_accounts`, Accounts & spend): Corporate, personal or unknown, with the organisation and email domain
- **Usage and spend** (`ai_usage`, Accounts & spend): Token and usage records the tools already wrote locally
- **Projects and data** (`ai_projects`, Projects & data): Repositories as owner/name, with a bounded data-file inventory
- **Device facts** (`device_facts`, The device): Make, model, OS and version
- **Fleet enrolment** (`mdm_enrollment`, The device): Whether the device is enrolled in management
- **Automatic updates** (`auto_updates`, The device): Whether OS updates install without being asked
- **Patch level** (`os_patch_level`, The device): How far behind the operating system is
- **Disk encryption** (`disk_encryption`, Security posture): Whether the volume is encrypted
- **Screen lock** (`screen_lock`, Security posture): Whether the screen locks, and after how long
- **Firewall** (`firewall`, Security posture): Whether the host firewall is on
- **Endpoint protection** (`edr_present`, Security posture): Whether an EDR agent is present
- **Network posture** (`network_posture`, Security posture): The device's network configuration
- **Browser policy** (`browser_policy`, Security posture): Whether browser policy is in force. Verified, never modified
- **Password policy** (`password_policy`, Security posture): Whether a password policy is enforced
- **Admin accounts** (`admin_accounts`, Security posture): Which accounts on the device hold administrator rights

## Blog

Posts by the endlayer team, newest first. Each is a fact the agent already measures on real devices; none claims a capability the product pages do not. RSS: https://www.endlayer.ai/feed.xml.

- [Introducing endlayer — the control plane for AI](https://www.endlayer.ai/blog/introducing-endlayer) — 2026-09-09, Product, The endlayer team. endlayer is an AI control plane for the devices people actually work on. One small agent, without administrator rights, reads 21 signals from files already on each machine and starts reporting within 60 seconds of install, and three modules turn those readings into a control: Surface builds the AI inventory, Monitor watches what each tool reaches and costs, and Govern grades every device against nineteen versioned rules.

## Frequently asked

### Does it read our conversations?

Yes, for PII leaking into a prompt, and for tools being used in ways nobody authorised. It is not on by default, and turning it on records who enabled it and the reason why. Your admin can narrow what is collected at any time, and the agent cannot widen it.

### Do we have to put a proxy or gateway in the path?

No. Nothing changes on your network. Everything is read from files already on the disk, which is also why desktop apps, coding agents and local models show up at all. none of them go through a gateway.

### Does the agent need admin rights?

No. It runs as the person using the machine and reads only what that account can already read. It never elevates, and it never executes any of the AI tools it finds.

### What if we use a tool that is not on your list?

Seventeen are recognised by name. Anything else still surfaces as an unidentified AI process with the account and the machine attached, and naming it properly takes a signature, not an integration, so there is no per-tool work for you either way.

### How quickly does something new appear?

Within fifteen minutes. Every device sends a snapshot every fifteen minutes, so a tool installed at 9:00 is in the inventory by 9:15.

### Can it block anything, or does it only watch?

Both. Every signal is measured against your policy version, and what fails gets an owner, a state and a date it closed. Real-time guardrails inside the agent act on the device as it happens: prompts are scored before they run, and secrets and personal data are warned on, redacted or blocked on the way out. Removing an installed app is the one action that waits for a person: an administrator approves the request before anything is uninstalled. Every action is recorded.

## How to cite this site

Cite the specific page URL, not just the domain. Canonical URLs are absolute and rooted at https://www.endlayer.ai. Attribute the product to endlayer and the company to Holistic AI.

## Optional

- Crawling policy: see [robots.txt](https://www.endlayer.ai/robots.txt). AI crawlers are permitted; attribution to the source URL is expected.
- Structured data: every page carries a Schema.org JSON-LD graph (Organization, WebSite, WebPage, and FAQPage on the homepage).
- Last generated: 2026-09-12