QuilrAI
Solutions · Employee AI Use

What your people
use at work.

Coding agents, chat tools, vendor copilots, self-hosted models on dev laptops. QuilrAI discovers all of it, governs every prompt and tool call, and gives security an audit trail without slowing engineering down.

100%

Endpoint coverage

macOS · Windows

150+

MCP tools governed

1 URL · per-agent scope

<40ms

Inline overhead

p99 across stages

0

SDK changes

in your IDE / browser

Scenario 01

Coding Agents

AI tools that read your repos, write code, run shells, and call MCPs.

CursorClaude CodeGitHub CopilotWindsurfDevin

The pain

An engineer pastes the AWS prod key into Cursor.

Cursor reads .env, opens config files, makes a tool call to an unfamiliar endpoint. The credential leaves the laptop before security sees the request.

READ ./.env → DATABASE_URL, AWS_SECRET
POST api.unknown-service.io/v1/... → JWT secret in payload
BLOCKED · scope violation · alert sent to security

Threats Blocked

Secret exfiltrationBash scope abuseRepo-wide readUndeclared network callsSkill chain misuse

Quilr Surface

Endpoint Agent · MCP Gateway · Guardian Agent

TLS-inspecting endpoint agent + per-tool MCP scope. No SDK changes in IDE.

Scenario 02

Chat Tools

Public AI services your team uses for daily work.

ChatGPTClaude.aiGeminiPerplexity

The pain

A finance lead pastes the Q3 forecast into ChatGPT.

MNPI leaves the company. No DLP fires. No audit trail. Compliance finds out 90 days later when the model gets retrained.

PROMPT chat.openai.com → 'Summarize this Q3 forecast: ...'
DETECTED · MNPI keywords + dollar amounts + entity names
REDACTED · prompt rewritten · user coached, request allowed

Threats Blocked

MNPI / PII / PHI leaksPrompt injection (return path)Shadow accountsNo audit trailRegulatory exposure

Quilr Surface

Endpoint Agent · Browser Extension · Compliance API

Browser extension scans every form post. Endpoint agent inspects TLS. Compliance API syncs Claude.ai org data.

Scenario 03

Vendor Copilots

AI built into the productivity suites your company already pays for.

Microsoft 365 CopilotGoogle Workspace AI

The pain

Copilot in Word reads the entire SharePoint, including the legal team's drafts.

Vendor copilots inherit the user's permissions across the whole tenant. They surface content the user has access to but shouldn't have summarized to a prompt window.

QUERY copilot.microsoft.com → 'Draft an email about the lawsuit'
RAG SharePoint /legal/drafts/*.docx → 12 documents fetched
BLOCKED · sensitivity label "Legal-Privileged" · request denied

Threats Blocked

Cross-tenant data exposureSensitivity-label bypassPermissions inheritance abuseRAG-side leakNo audit ownership

Quilr Surface

Browser Extension · Endpoint Agent · Quilr SDK

Browser extension layered on top of Microsoft 365 / Workspace. Endpoint agent catches what's in transit.

Scenario 04

Self-Hosted on Dev Machines

Open-source models and community MCPs running on engineering laptops.

OllamavLLMCommunity MCPs

The pain

An engineer runs Ollama with a community MCP server. Nobody in IT knows.

Local models bypass every proxy. Community MCPs ship with broad scopes by default. Discovery is the first problem; runtime governance is the second.

DISCOVER ollama.ai/api/generate listening on localhost:11434
MCP @community/file-system → unrestricted disk read enabled
ENROLLED · scope tightened · usage telemetry enabled

Threats Blocked

Shadow AI on dev machinesUnrestricted MCP scopesLocal model bypassNo telemetryUntracked exfil channels

Quilr Surface

Endpoint Agent · MCP Gateway · Discovery API

Endpoint agent enumerates local LLM endpoints + community MCPs, ties each to a Guardian for runtime enforcement.

How it plugs in

One platform · four touchpoints.

Same Decision Engine. Different surfaces depending on where the AI tool lives — on the laptop, in the browser, in the agent.

Endpoint Agent

TLS inspection + DLP on macOS and Windows. Catches every AI request leaving the laptop — no kernel extension required.

Browser Extension

Scans form posts, prompts, and pastes in real time. Works for ChatGPT, Claude.ai, M365 Copilot, Workspace AI.

MCP Gateway

One URL routes every MCP tool call. Per-agent identity, per-tool scope. Auto-detects Cursor / Claude Code / OpenAI / Gemini.

Compliance API

Syncs orgs, users, chats, and DLP findings from Claude.ai and ChatGPT Enterprise. Audit evidence on demand.

Compliance coverage

Audit evidence on day one.

Every prompt, redaction, and block produces a per-user log entry mapped to the framework your auditor cares about.

SOC 2 Type II

Audit trail per user · DLP evidence · access reviews

HIPAA

PHI detection · redaction · BAA-ready logs

NIST AI RMF

Map · Measure · Manage · Govern across all AI tools

EU AI Act

High-risk system documentation · transparency obligations

Get started

See it on your own AI inventory.

30-minute walkthrough. We'll discover what's running on your endpoints, create Guardian Agents, and run a live red team — on your stack, not a demo dataset.