AgentLens

See what your AI agents can access.

Scans your machine for installed AI coding agents, maps their data source connections, and flags hardcoded secrets in config files.

Install

npx nox-agent-lens

# Or install globally
npm install -g nox-agent-lens

What it detects

Agents — Claude Code, Claude Desktop, Cursor, Windsurf, GitHub Copilot, Continue, Cline, Amazon Q

MCP Connections — Per-agent data sources configured via MCP (Model Context Protocol), including sub-services and transport details

Hardcoded Secrets — API keys, tokens, and credentials stored as plaintext in MCP config files instead of environment variable references

Shell Access — CLI tools installed via Homebrew and npm that any agent can invoke

Usage

# Terminal UI
agent-reveal

# JSON output (for automation)
agent-reveal --json

# Fleet aggregation — merge scans from multiple machines
agent-reveal aggregate scans/*.json

# Fleet aggregation — JSON output
agent-reveal aggregate --json scans/*.json

# Fleet aggregation — export to file
agent-reveal aggregate --out report.json scans/*.json

Fleet workflow

# 1. On each machine, export a JSON scan:
agent-reveal --json > /shared/scans/$(hostname).json

# 2. Aggregate into a fleet-wide report (terminal dashboard):
agent-reveal aggregate /shared/scans/*.json

# 3. Or export the report as JSON:
agent-reveal aggregate --out fleet-report.json /shared/scans/*.json

Example output

Single machine scan

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  AgentLens
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  eng-macbook-01 · darwin arm64 · 3/9/2026

  ┌──────────────────────┐  ┌──────────────────────┐  ┌──────────────────────┐
  │ 👾  Claude Code      │  │ 🧊  Cursor           │  │ ✴️  Claude Desktop   │
  │ 1.0.78               │  │ v0.48.9              │  │ v4.0.3               │
  │ 3 MCP                │  │ 4 MCP                │  │ 2 MCP                │
  │ 🚨 2 secret(s)       │  │ 🚨 4 secret(s)       │  │                      │
  └──────────────────────┘  └──────────────────────┘  └──────────────────────┘

▸ 👾  Claude Code — MCP Connections
    ✔ gws  (drive, gmail, calendar, sheets, docs, chat)
    🚨 Datadog MCP
       plaintext credentials: DD_API_KEY, DD_APP_KEY

▸ 🧊  Cursor — MCP Connections
    ✔ gws  (drive, gmail, calendar, sheets, docs, chat)
    🚨 Datadog MCP
       plaintext credentials: DD_API_KEY, DD_APP_KEY
    🚨 Jira MCP
       plaintext credentials: ATLASSIAN_API_TOKEN, ATLASSIAN_EMAIL

▸ ✴️  Claude Desktop — MCP Connections
    ✔ github  (repos, issues, pull-requests)
    ✔ linear  (issues, projects)

▸ 🐚  Shell Access
  All 3 agents share this machine's CLI tools.
  brew (13): gh, terraform, redis, mongodb, ...
  npm (17): @railway/cli, vercel, netlify-cli, ...

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Exposure Summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  3 AI agents on this machine can access:

    5 data sources via MCP
    30 CLI tools via shell

  🚨 6 hardcoded secrets in 2 data source(s)
  Plaintext API keys and tokens sitting in JSON config files.
  Any agent — or anyone with access to this machine — can read them.

  Recommendation: Use ${ENV_VAR} references instead of raw values.

Fleet aggregation

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Fleet Report
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  5 machines · 2026-03-09

▸ Machines
  ✔ data-linux-01         Linux     1 agent     3 MCP     🚨 2 secrets
  ✔ devops-mac-01         Darwin    5 agents    10 MCP    🚨 1 secret
  ✔ eng-macbook-01        Darwin    4 agents    6 MCP     🚨 1 secret
  ✔ eng-macbook-02        Darwin    2 agents    4 MCP
  ✔ intern-macbook-01     Darwin    1 agent     1 MCP

▸ Agent Coverage
  👾  Claude Code       4/5 machines   (devops-mac-01, eng-macbook-01, ...)
  🧊  Cursor            3/5 machines   (data-linux-01, devops-mac-01, eng-macbook-01)
  🏄  Windsurf          1/5 machines   (devops-mac-01)
  🤖  GitHub Copilot    1/5 machines   (eng-macbook-01)
  🔄  Continue          1/5 machines   (eng-macbook-01)
  🔧  Cline             1/5 machines   (eng-macbook-02)
  📦  Amazon Q          1/5 machines   (devops-mac-01)
  ✴️  Claude Desktop    1/5 machines   (devops-mac-01)

▸ Data Sources — 14 found
  datadog       3 machines  👾 ✴️ 🧊 🏄  MCP @datadog/mcp-server
                dashboards, logs, metrics, monitors
  github        3 machines  👾 🧊 🤖 🔧  MCP @modelcontextprotocol/server-github
                issues, pull-requests, repos
  aws           1 machine   👾 🧊 🏄 📦  MCP @aws/mcp-server
                cloudwatch, ec2, iam, lambda, s3
  docker        2 machines  👾 🧊        CLI docker
                containers, images, volumes
  postgres      1 machine   👾 🔄        MCP @modelcontextprotocol/server-postgres  🚨
  linear        1 machine   👾 🔧        MCP @linear/mcp-server
                issues, projects
  ...

▸ Hardcoded Secrets — 4 found
  🚨 data-linux-01         snowflake     SNOWFLAKE_TOKEN
  🚨 data-linux-01         s3            AWS_SECRET_ACCESS_KEY
  🚨 devops-mac-01         pagerduty     PAGERDUTY_TOKEN
  🚨 eng-macbook-01        postgres      POSTGRES_PASSWORD

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Fleet Summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  5 machines scanned:

    14 agent installs across 8 tools
    14 unique data sources via MCP & CLI
    4 hardcoded secrets on 3 machines

  🚨 Most exposed: devops-mac-01

  Recommendation: Use ${ENV_VAR} references instead of raw values.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Detailed Inventory
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  HOSTNAME           AGENT           CONNECTION  PROTO   SUB-SERVICES                       SECRETS              RISK
  ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────
  data-linux-01      Cursor          snowflake   MCP     warehouses, databases              SNOWFLAKE_TOKEN      🚨 critical
  data-linux-01      Cursor          s3          MCP     buckets                            AWS_SECRET_ACCESS_KEY 🚨 critical
  devops-mac-01      Claude Desktop  pagerduty   MCP     incidents, schedules               PAGERDUTY_TOKEN      🚨 critical
  eng-macbook-01     Claude Code     postgres    MCP     —                                  POSTGRES_PASSWORD     🚨 critical
  devops-mac-01      Amazon Q        aws         MCP     ec2, s3, lambda, iam, cloudwatch   —                    ⚠️ medium
  devops-mac-01      Claude Code     kubernetes  CLI     pods, deployments, services, logs  —                    ⚠️ medium
  eng-macbook-02     Cline           github      MCP     repos, issues, pull-requests       —                    ⚠️ medium
  ...

MDM deployment (JumpCloud, Jamf, etc.)

Run across all endpoints via your MDM command center:

CURRENT_USER=$(stat -f "%Su" /dev/console)
NPX=$(find /usr/local/bin /opt/homebrew/bin -name npx -maxdepth 1 2>/dev/null | head -1)
su - "$CURRENT_USER" -c "$NPX -y nox-agent-lens --json" 2>/dev/null

Requirements

• Node.js 18+
• macOS (Linux and Windows support planned)

License

MIT