Lumbra
by Lumbra Inc.
LumbraOS: governed, auditable AI agent orchestration for classified and regulated environments
Lumbra makes LumbraOS, a model-neutral platform for running AI agents where every output must be traceable to its source. It is built for defense and intelligence agencies and for regulated firms in finance, healthcare and law that need audit logs, evaluation gates and human sign-off before an agent writes to a system of record.
Lumbra is a Washington, DC company founded in 2025 by CEO Aaron Brown, a former Army Ranger and CIA counterterrorism officer, and CTO Michael Hiebert, a former Army Ranger and MIT-trained computer scientist. Its product, LumbraOS, orchestrates AI agents in what the company calls high-consequence and ultra-secure computing environments. It has three components. Nebula runs agents, logs their actions, applies data protections and traces each agent claim back to its primary source so a domain expert can validate it before the finding reaches a system of record. StarChart is a customer-specific knowledge graph where agents accumulate insights alongside their sources and embedded sensitivity labels. Astrolabe benchmarks agents before deployment and monitors them in production against the customer's own benchmarks, with in-line checks for hallucination and drift. The platform is model-neutral and can mix frontier and open-weight models; Unite.AI reports support for air-gapped, on-premises and sovereign-cloud deployment. Lumbra's agents run in production on top-secret networks across the Pentagon and the US intelligence community. It holds a multimillion-dollar contract with an unnamed intelligence agency, and the Pentagon's CDAO selected it alongside Palantir for the Agent Network pace-setting project. On November 10, 2026 it announced a $25 million Series A co-led by Point72 Ventures and Ansa Capital, with seed investor Costanoa Ventures, to expand into financial research, KYC/AML, sanctions and supply-chain risk, legal and clinical workflows.
A CIO or chief data officer at a defense, intelligence or regulated financial organisation who must deploy agents on sensitive networks and prove where every output came from.
Agent findings arrive with a provenance chain, evaluation results and a human approval step, so they can pass audit and security review.
At a Glance
- Category
- AI Agents & Orchestration
- Pricing
- Contact for pricing
- Target Market
- CIOs, CTOs, Government agencies, Financial services, Healthcare
- Deployment
- Self-hosted, Hybrid
- Founded
- 2025
- Headquarters
- Washington, DC, United States
Key Features
- ✓Nebula orchestration
Runs multi-agent workflows, logs every action and links each claim to its primary source, giving reviewers a full provenance chain.
- ✓StarChart knowledge graph
Builds a customer-specific knowledge graph where insights are stored with their sources and sensitivity labels, so knowledge compounds across tasks.
- ✓Astrolabe evaluation
Benchmarks agents before deployment and monitors them in production for hallucination, drift and confidence collapse against customer-defined benchmarks.
- ✓Human validation gate
Requires a domain expert to approve findings before agents write to a system of record, keeping accountability with people.
- ✓Model-neutral runtime
Mixes proprietary frontier models and open-weight models per task, so customers can choose by performance and by what their network allows.
- ✓Secure deployment options
Supports air-gapped, on-premises and sovereign-cloud setups and already runs on top-secret defense networks, per company statements.
Capabilities
Use Cases
- •Intelligence analysis
Analysts at a US intelligence agency run agents on classified networks to assemble source-linked findings that experts verify before use.
- •Defense agent network
The Pentagon's CDAO uses Lumbra for AI orchestration in its Agent Network pace-setting project, alongside Palantir.
- •Sanctions and KYC/AML research
Financial-data providers and research teams use agents for diligence, sanctions screening and supply-chain risk with auditable sourcing.
- •Clinical workflow support
Healthcare engagements include live surgical-preparation workflows, with revenue-cycle and population-health work named as adjacent uses.
Ideal For
Best For
- ✓Intelligence and defense analysis on classified or air-gapped networks
- ✓Financial due diligence, sanctions screening and KYC/AML research that needs source-linked findings
- ✓Supply-chain risk analysis across many datasets with different sensitivity levels
- ✓Organisations that must evaluate and monitor agents continuously against their own benchmarks
- ✓Teams mixing frontier and open-weight models under one governance layer
Not Ideal For
- ✗Small teams wanting a self-serve agent builder with public pricing; Lumbra sells through enterprise and government contracts
- ✗Fully autonomous, unsupervised workflows; the design requires a domain expert to validate outputs, and the CEO has said agents are not ready for high-stakes decisions such as targeting
- ✗Buyers who need published FedRAMP, SOC 2 or similar attestations up front, since none are listed on the website
Deployment
Market Analysis
Pros
- ✓Proven in the most demanding security setting: production use on Pentagon and intelligence-community top-secret networks
- ✓Provenance, evaluation and human approval are built into the core workflow
- ✓Model-neutral, so customers are not tied to one model vendor
- ✓Founders bring operational intelligence and special-operations backgrounds that match its core buyers
Cons
- ✗Young company (founded 2025) with commercial traction in finance and healthcare still early and mostly unnamed
- ✗Most traction claims come from the company; the funding coverage is largely promotional and offers no independent verification
- ✗No public pricing, documentation, SDK or listed compliance certifications, and no G2, Hacker News or GitHub presence to judge practitioner experience
Pricing
LumbraOS
Contact for pricing
- ✓Nebula orchestration
- ✓StarChart knowledge graph
- ✓Astrolabe evaluation
- ✓Air-gapped, on-prem or sovereign-cloud deployment
Lumbra publishes no pricing. It sells through government contracts (one described as multimillion-dollar) and enterprise B2B deals, so cost depends on deployment environment and scope.
Security & Compliance
Connect
Sources
This page was written from 4 sources, 3 on domains other than lumbra.ai.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
StackAI
No-code enterprise AI agent builder with VPC and on-premises deployment, now owned by Asana
GetVocal
Governed hybrid human-AI agents for enterprise contact centers in regulated industries
NeuBird Agentic Reliability Center
AI SRE agent and governed control plane for every AI agent that touches production
Traversal
Enterprise AI SRE for alert triage and root cause analysis across existing observability tools