The Governed Agentic AI Platform

The Legion Intelligence platform turns mission intent into finished work across the systems and data you already run. It runs at IL2 through IL6, FedRAMP High in the cloud, and fully air-gapped. Agents carry out the work. Authority stays with command. Every action is attributed and audited.

In use across

  • U.S. Department of War
  • U.S. Department of Energy
  • U.S. Special Operations Command
  • U.S. Naval Special Warfare Command
  • U.S. Marine Corps Forces - Special Operations Command

Getting an answer is not getting the work done.

Three gaps sit between an AI answer and a completed operational task.

01

Insight without action

Chat tools answer questions. A person still does every step after that.

02

Capability without control

AI expands what a team can do. Without governance, it expands what can go wrong.

03

Tools without mission reality

Enterprise software assumes connectivity, clean data, and an unclassified network. Operations rarely offer all three.

What is a governed agentic AI platform?

Legion is a governed agentic AI platform: AI agents complete multi-step work, and each consequential step waits for a person with the authority to approve it. Teams use it for intelligence production, operational reporting, and maintenance coordination.

What runs underneath.

Download Product Sheet

Operational work breaks the assumptions ordinary software is built on. Several people and agents share one product, classified data has to keep its markings, consequential steps wait for someone with authority, and the network can drop for days. The six capabilities below handle all of that at once, in every environment Legion runs in.

Governed Capabilities

People and agents on one shared document

Multiplayer runs a recurring staff process as one shared document, open to authorized staff and scoped agents at the same time. Every change is attributed, gated steps wait for an authorized reviewer, and the finished product carries its approvals, sources, and record of who changed what. Agents can propose a classification marking. They never set one.

Your team builds and changes its own apps

The people who own the work describe it in plain language, and Legion builds a working app from governed parts that already carry permissions, classification handling, and audit. Teams start from the closest Pack and reshape it the week requirements change. A services customer now drafts work orders in 90 seconds that used to take hours.

Answers from the data you already hold

Agents reason across documents, records, and message traffic where they already sit, with no migration into a new schema first. Every answer cites its source and carries that source’s classification markings, and 100+ integrations, including IBM Maximo, ServiceNow, and SharePoint, keep your existing permissions intact.

Nothing moves without your name on it

Permissions decide what an agent can see. Checkpoints decide what it can do, so a report, a task route, or a record update waits for the person with the authority to approve it. Any action can be paused or overridden, and every step lands in an append-only record that exports as evidence for an ATO package or IG review.

Every model tested on your own work

The Eval Loop scores models, agents, and workflows against the tasks your teams perform, using measures your organization defines. Runs stay inside your boundary, connected or air-gapped, and every new model is re-baselined against your work before it takes over a task, so drift gets caught before it reaches operations.

The right model for each task

Frontier, open-weight, and domain-specific models run side by side behind one gateway, so a routine summary doesn’t pay frontier rates. New models plug in without code changes as approvals shift, spend follows the roles you already administer, and the same controls govern cloud APIs and the models running on Centurion at the edge.

Model Supply

  • Frontier Models

    Customer-approved, connected deployments.

  • Open weight models

    Locally-hosted, on-prem, air-gapped.

  • Domain-specific models

    Fine-tuned for the task.

Deployment Environments

  • Cloud

    FedRAMP High

  • On-prem

    No cloud dependency

  • Air-gapped and classified

    IL2 through IL6

  • Edge

    Centurion, DDIL

Integrate With the Systems You Use Every Day

  • Salesforce
  • Oracle
  • S.A.P.
  • Zendesk
  • Slack
  • Siemens
  • Microsoft Sharepoint
  • I.B.M.
  • Microsoft Outlook
  • ...and more.

Three ways to put Legion to work.

Legion Packs, Centurion by Legion Intelligence, and Mission Builder share one platform.

Packs

Choose Packs, not prompts.

Role-specific AI apps out of the box, with humans in command. Packs are how Legion is purchased and used.

Learn More
centurion

Cloud when you can. Edge when you must.

Centurion packages Legion into a deployable system for DDIL, classified, and edge environments. Same Packs, same governance.

Learn More
Builder

Build the tools your work needs.

Teams can shape what they need from governed parts. No engineering queue.

Learn More

Built for classified, air-gapped, and degraded environments.

Legion is built for the conditions high-consequence work actually runs in: classified, air-gapped, degraded, and disconnected environments. It runs on classified networks at IL2 through IL6 and deploys fully disconnected. Moving a deployment into a classified or disconnected network requires no re-architecture: what a team builds in one carries to the next.

Centurion is the deployable edge system integrating hardware, software, and sustainment that runs Legion in DDIL (denied, degraded, intermittent, or limited connectivity) conditions. Agents, workflows, and models run locally on the deployed system: Legion Packs operate offline-first, hold work through disruption, and synchronize when connectivity returns and policy permits. Governance does not loosen when connectivity drops.

Proven where the work happens.

Legion is in use across the U.S. Army, U.S. Air Force, the U.S. Department of Energy, the Missile Defense Agency, and the Defense Industrial Base. Same platform that was proven by 20,000+ users across U.S. Special Operations Command.

  1. 48x Faster

    INTSUM generation at XVIII Airborne Corps

  2. 216 hours saved per weekly

    Across USSOCOM battalion staff workflows

  3. 88% Faster SITREPs

    Reduced from 8 hours to under 1 hour

  4. 25% More Technician Capacity

    With the Maintenance Technician Pack

  5. Operational in 1 Day

    From field install to mission-ready (Scarlet Dragon 26-01)

Why not a chat tool, a custom build, or an integrator-led platform program?

A Chat Assistant

Chat tools answer questions, and a person still performs every step in between. On the Legion platform, Apps complete multi-step work under human approval, so what you measure is the job done.

A Custom Build

Internal builds spend months on identity, permissions, integrations, and auditability before the first workflow reaches users, and each new environment becomes a separate project. Legion arrives with governance, deployment patterns, and integrations already in place, and adapts as the work changes.

An Integrator-led Platform Program

These programs deliver on the integrator's timeline: months of services before a first outcome, changes routed through a vendor backlog, and a stack standardized on that vendor's models and terms. Legion deploys in weeks and adapts in days, by your teams, with the model choice staying yours.

Security for classified and
high-consequence environments.

Legion brings security designed for the U.S. Department of War to every deployment. The platform enforces zero-trust principles to protect data, secure access, and contain the impact of any breach through segmentation and policy enforcement. Accreditation and security posture:

  • FedRAMP High
  • IL2 through IL6
  • SOC 2 Type II certified
  • NIST 800-53 compliant
  • Ready to manage ITAR data
  • CMMC 2.0 Level 2
  • Air-gapped deployment capable
  • HBOM and SBOM available for accreditation

See the platform in your environment.

See how Legion would run for your team: which Packs fit your roles, which systems connect first, and how a pilot expands into a larger footprint.

FAQs

What is the Legion platform?

The Legion platform is a governed agentic AI platform for environments where security, reliability, and accountability are non-negotiable. AI agents complete multi-step operational work (intelligence production, operational reporting, maintenance coordination) under human authority, across the systems and data you already run, with every action attributed and audited.

How is the platform different from a chatbot or AI assistant?

A chatbot answers questions while a human still does every task. The Legion platform runs Legion Packs whose Apps carry out the work directly: drafting products, coordinating across systems, and updating records, all under human approval with full attribution. The completed job is what the platform delivers.

What are Legion Packs, and how do they relate to the platform?

Legion Packs are role-based capability bundles that deploy a ready Space with the right Apps. The platform is what every Pack runs on: the models, integrations, governance, and deployment underneath. Centurion runs the same platform at the edge, so Legion Packs and Centurion share one foundation rather than being separate products.

Which AI models does the platform support?

The Legion platform is model-agnostic. It runs open-source, proprietary, and fine-tuned models, routing to the latest customer-approved frontier models in connected deployments and running locally hosted open-weight models in air-gapped and on-prem deployments. New models are added through the Model hub without code changes, and you choose the best model for each task.

How does Legion test an AI model before it takes over a task?

The Eval Loop scores models, agents, and workflows against the work your teams perform, using measures your organization defines in code. Evaluation runs inside your own boundary, on connected and air-gapped networks, under the requesting user’s identity, with every run logged. Each new model is re-baselined against your work before it takes over a task.

Can the platform run without cloud connectivity?

Yes, through the right deployment. On-prem and air-gapped deployments run with no cloud dependency, with models hosted locally. For DDIL (denied, degraded, intermittent, or limited connectivity) conditions, Centurion by Legion Intelligence runs the same Packs and governance on a deployable edge system, offline-first, synchronizing when connectivity returns.

What environments and classification levels does it support?

The platform deploys across cloud, private cloud, hybrid, bare metal, on-prem, and air-gapped and classified networks, and runs at IL2 through IL6, including classified SECRET environments. The same Packs and governance run identically across all of them, with no re-architecture required to move between them.

How does Legion keep humans in control of AI agents?

Operators define objectives and constraints, agents carry out the steps, and every action runs inside policy you set. Human-in-the-loop checkpoints sit at each stage, and any authorized user can pause, redirect, or override an action. Every action is logged with full attribution, including agent actions.

How do people and AI agents work on the same document in Legion?

Legion Multiplayer runs a recurring staff process as one shared document, open to authorized staff and scoped agents at the same time. Every change is attributed to the person or agent who made it, and gated steps wait for an authorized reviewer. Agents can propose classification markings, while only authorized people set them. Multiplayer runs on SIPR and at the disconnected edge.

What systems does the platform integrate with?

The platform connects to more than 100 systems out of the box, including IBM Maximo, ServiceNow, Microsoft SharePoint, and Microsoft Teams, with MCP server support for agent tools. Legacy and custom systems of record connect through the same API framework, with your existing permissions enforced end to end.

Does Legion require migrating data into a new system before it’s useful?

No. Legion’s Dynamic Knowledge Engine reasons across documents, structured records, and message traffic where they already live. Classification markings and provenance travel with every passage, so each answer cites an authoritative source. Legion connects to 100+ systems, including IBM Maximo, ServiceNow, and Microsoft SharePoint, and enforces your existing permissions end to end.

What security and accreditation does the platform hold?

The Legion platform is FedRAMP High and operates at IL2 through IL6. It is SOC 2 Type II certified, NIST 800-53 compliant, and ready to manage ITAR data, with air-gapped deployment capability and HBOM and SBOM available for accreditation. Your data stays in your environment and never trains external models.

How quickly can the platform be deployed?

The Legion platform has been installed and operational in one day in field evaluations, and standard deployments are scoped at 30 days. It connects to existing systems and data without migrations or infrastructure replacement, so teams start running governed work quickly.

What is Mission Builder?

Mission Builder is how customers and partners build their own Packs. Teams describe what they need in plain language, build a Pack tuned to their requirements, and extend or modify the Packs Legion ships as the work changes. Every Pack built in Mission Builder inherits the platform's permissions, classification handling, and audit.