Your environment, your models
Runs on your hardware or in your own cloud account, with models inside your boundary or a provider you already use. No connection back to Decision Terrain.
Review deployment optionsGovernment & Public Sector
AI agents for government programs: persistent agents for program reporting, document review, and administrative coordination, with a person approving what leaves the system.

Public-sector operating constraints
Customer-controlled options
Optional DT engineers
01 / Useful agent work
Agents keep their instructions, files, and history. Teams assign an ongoing responsibility, inspect progress, and review the result together.
Illustrative applications of supported capabilities, not preconfigured mission systems or claims of deployed customer results.

02 / Why security says yes
Most agent projects stall at the security gate, not on capability. Decision Terrain is designed around the three questions an approver asks first: where does the data go, can we stop it, and can we prove what it did.
Runs on your hardware or in your own cloud account, with models inside your boundary or a provider you already use. No connection back to Decision Terrain.
Review deployment optionsDeny tools for one agent, a Team, or everyone. Hold outgoing email for review of the exact message, and require approval for exact API writes. Instructions guide agents; these controls enforce boundaries.
Review security and governanceWhere enabled, place a security hold from the console or from your SIEM, released only by an authorized person. Material actions are recorded before they run and forwarded to your security tools.
Request the security review materials03 / Operating constraints
Data boundaries, authorities, identity, model access, and connectivity decide what can be useful in a public-sector environment.
Your hardware or your own cloud account, including a supported air-gapped configuration.
Review deployment optionsConnect approved internal APIs and MCP tools with explicit credentials, and bring agent work into your applications.
Explore integration optionsRole-based administration, activity records, and configured review for covered actions.
Review security and governanceGive evaluators the available audit evidence and capture limits. Map security-event delivery to the approved topology; customer hosting alone does not make a public-cloud route disconnected.
04 / Getting started
The platform is built by people with intelligence and defense experience and contributions to Browser Use and LiteLLM. DT engineers are available for setup.
Build with platform documentation, or add DT engineers for setup and integration.
Engineering helpAir-gapped deployment, internal integrations, and relevant engineering experience.
Explore defense and intelligenceMission and contributor experience does not imply agency sponsorship or project endorsement. Customer security and authorizing teams retain their decisions and authorities.
05 / A practical first step
Bring an unclassified, non-sensitive description of the work, users, information, and constraints. We'll shape a demonstration around it.
How agents, connected systems, shared workspaces, and reviewable deliverables fit together.
Explore the platformGuides and planning tools for readiness, evaluation criteria, and a bounded test.
Visit the resource centerNext steps
Explore the platform around your systems and workflows. Your team can lead implementation, with DT engineering and support available when you need it.