Gotham

Gotham is Palantir Technologies’ intelligence and operations platform, first deployed in 2008 and built to give analysts a single working surface for fusing classified sensor feeds, signals intelligence, human-source reporting, and open-source material. It emerged from Palantir’s early CIA-adjacent contract work — In-Q-Tel was an initial investor — and was designed for the counterterrorism analyst’s core problem: tens of thousands of fragmentary records about people, places, vehicles, and events scattered across systems that could not talk to each other.

At its heart Gotham is entity-centric. Records from each source are resolved into objects — a person, a vehicle, a building, a phone number, a transaction — and every relationship between them is preserved as a typed link. Analysts work the resulting graph through link-analysis tools, a geospatial layer that places those entities on a map, and a time scrubber that shows how the network evolved. Newer releases fold in machine learning for pattern-of-life detection, anomaly flagging, and target development, and the platform now exposes Palantir’s AIP — its agentic LLM layer — so operators can interrogate the data set in natural language and chain inference steps across structured and unstructured sources.

The platform is fielded across the US intelligence community and the Department of Defense, with publicly disclosed contracts running into the hundreds of millions of dollars: the Army’s Distributed Common Ground System replacement, the TITAN ground-station programme, and a string of Marine Corps and Special Operations Command awards have folded Gotham in over the last decade. Allies running it include the United Kingdom, Israel, France, and — most visibly since 2022 — Ukraine, where it underpins targeting workflows in the war with Russia. Palantir CEO Alex Karp has said publicly that the company’s software handles most of the targeting work on the Ukrainian side, a claim broadly echoed by Ukrainian commanders and Western reporters embedded with the war effort.

Gotham itself has been continuously iterated rather than re-versioned. The principal recent shift has been the AIP layer announced in 2023, which moved the platform from a query-and-graph tool to an agentic workspace where operators describe an effect they want and the system proposes courses of action against the connected sensor and weapons feeds. The same backbone now feeds into the Maven Smart System work the company runs for the Pentagon and into Project Convergence experiments. A lighter intelligence variant called MetaConstellation appeared briefly in 2022, pitched at small allied governments, before folding back into the main product line.

Among intelligence platforms Gotham occupies the high end: more analyst-facing than the BAE and Raytheon ground systems that compete with it on raw ingest, and more deeply wired into operational targeting than visualisation tools such as IBM i2. That positioning, more than any single feature, is what has kept it inside the targeting cycles of the largest Western militaries for nearly two decades.