Neo

Neo is an endpoint-based control platform that finds the AI agents, AI-enabled applications, MCP servers, plugins and extensions running on employee devices, assesses what each can access, and enforces policy on their actions in real time. It is built for enterprise security teams that need to govern agentic software without banning it.

Neo, from Boston-based Neo Security, is an agentic software control platform that came out of stealth on July 20, 2026 with $100 million from Andreessen Horowitz and Bessemer Venture Partners, with Craft Ventures and Merlin Ventures participating. The company was founded in 2025 by Nick Warner (CEO, formerly president and COO of SentinelOne), Shlomi Salem (CPO, formerly a SentinelOne research and detection engineering leader) and Eran Shirazi (CTO, co-founder of EasySend). The product runs as a 25MB endpoint sensor on macOS 12 and later, Windows 10 1607 and later, and Linux, without kernel modules or reboots, and the vendor says analysis is processed on the device. It works as a four-step loop. Discover inventories agents, plugins, MCP servers, extensions, models and ordinary enterprise software on every endpoint, backed by a knowledge base called Neoverse that Neo says maps more than 1.2 million artifacts. Assess evaluates each item's capabilities, permissions, configuration and attack paths. Attribute ties tool calls, file access and API activity back to the human or agent that caused them, producing an audit trail. Enforce applies runtime decisions (allow, block, redirect or hold for human approval) before a sensitive action executes. Policies can be written in plain language through a built-in LLM assistant, and an audit mode lets teams observe before enforcing. Supported software listed by the vendor includes Claude, ChatGPT desktop, OpenAI Codex, GitHub Copilot, Google Gemini, Cursor, Perplexity, Mistral Le Chat, Slack AI, Notion AI and Salesforce Agentforce. Neo has not published pricing or named customers. Its endpoint-first design separates it from gateway and proxy-based AI security tools.

Category

Governance & Security

Pricing

Contact for pricing

Target Market

CISOs, CIOs, Security Operations Teams, Endpoint Security Teams

Deployment

Edge-first, Cloud-first

Founded

2025

Headquarters

Boston, United States

• ✓
  Neoverse-powered discovery

  Continuously inventories agents, plugins, MCP servers, extensions, models and enterprise software on every endpoint, using a knowledge base the vendor says covers over 1.2 million artifacts.

• ✓
  Capability and risk assessment

  Evaluates what each discovered item can access, how it is configured and which attack paths it opens, so teams can find excessive privileges before an incident.

• ✓
  Human and agent attribution

  Links tool calls, file access and API activity to the specific person or agent that initiated them, creating an immutable audit trail for investigations.

• ✓
  Runtime policy enforcement

  Applies allow, block, redirect or require-approval decisions on the endpoint before a sensitive agent action executes, without needing a separate enforcement product.

• ✓
  Plain-language policy builder

  A built-in LLM assistant turns written rules into policy that deploys across all endpoints as a single rule set, lowering the effort to author agent controls.

• ✓
  Lightweight endpoint sensor

  A 25MB sensor for macOS, Windows and Linux that needs no kernel modules or reboots and, per the vendor, delivers a first inventory in under 15 minutes.

• ✓
  Audit mode

  Lets security teams watch what policies would do across the fleet before enforcement is switched on, reducing the risk of breaking developer workflows.

• •
  Securing AI coding assistants

  Control what agents such as Claude, Codex, Copilot and Cursor can read, execute and send off developer machines, so code assistants run without unchecked access to secrets and source.

• •
  MCP server governance

  Discover unmanaged MCP servers on endpoints, block unsigned ones and limit which tool calls are permitted, closing a channel most existing endpoint tools do not inventory.

• •
  Preventing agent-driven data loss

  Hold large data exports, flag credential exposure and restrict file access when an agent acts, stopping unauthorized data movement before it leaves the device.

• •
  Shadow AI inventory for audit

  Give security and compliance teams a current list of which AI tools, models and extensions each employee runs, with configuration and permission risk attached, ahead of audits or policy reviews.

• •
  Agent incident investigation

  Use attributed records of tool calls, file reads and API calls to determine whether a person or an agent performed an action during an investigation, shortening triage.

Neo publishes no list prices, plans or metering details as of October 2026. Access is through a sales demo, and neither the site nor launch coverage states whether it is priced per endpoint, per user or by platform license.

This page was written from 8 sources, 4 on domains other than neo.ai.

1. 1.neo.ai — neo.aivendor
2. 2.neo.ai — platformvendor
3. 3.neo.ai — neo launches 100mvendor
4. 4.neo.ai — aboutvendor
5. 5.securityweek.com — neo emerges from stealth with 100m to control and secure ent
6. 6.bankinfosecurity.com — neo launches 100m to guard agentic enterprise software a 322
7. 7.siliconsnark.com — neo raised 100 million to put a bouncer on your ai agents
8. 8.fintech.global — neo lands 100m as agentic ai outpaces security teams