Lumbra raises $25M Series A for AI agents in the intelligence community

Lumbra Inc. has raised $25 million in Series A funding to build autonomous agents that sort through sensitive intelligence data. Point72 Ventures and Ansa Capital led the round, and seed investor Costanoa took part. The company was founded last year by former CIA officer Aaron Brown and former special forces infantry officer Michael Hiebert.

Its platform, LumbraOS, is model-neutral, so operators can mix proprietary and open-weight models across different national security and intelligence tasks. It has three components: Nebula, which traces every agent claim back to its primary source data so experts can validate outputs; StarChart, a knowledge graph that links insights across datasets using sensitivity tags; and Astrolabe, which enforces governance policies and benchmarks and monitors agents. The founders say the system was built for classified computing environments.

The company says it won a multimillion-dollar contract with an unnamed U.S. intelligence agency, and the Pentagon named Lumbra as one of two companies underpinning its Agent Network project. Lumbra plans to use the new funds to expand its engineering and product teams and its go-to-market operations. It is also targeting enterprises in healthcare, finance and law, where it says traceable audit logs and governance controls can help agents run across larger networks.

Developers building agents for regulated or high-stakes settings can take note of the design patterns described here, including claim-level source tracing, audit logs and benchmarking agents before deployment. The round shows investors backing governance tooling for agents, not only general-purpose agent platforms. These capabilities are as described by the company and have not been independently verified.