A fort is not just a building that holds soldiers. In the Maratha state-building project, a fort was the basic unit of political control — a distributed system of sovereignty encoded in geography. Shivaji's 60+ forts were not defensive installations waiting for attack; they were the nodes of an administrative and military network that made the Maratha state persistent even when no Maratha army was in the field.
The clearest way to understand this: the Mughal empire collapsed into factionalism after Aurangzeb's death partly because its power was centralized in court hierarchies and mansabdar networks. The Maratha state survived multiple crises after Shivaji's death because its power was distributed across stone. You cannot bribe a fort into switching sides the way you can bribe a mansabdar. You have to take it militarily, which is slow, expensive, and uncertain in mountain terrain.
Every Maratha fort was administered by three officers — the havaldar (commandant), the sabnis (accountant), and the sarnobat (military commander) — who held authority in separate domains and were deliberately drawn from different communities and social backgrounds. The keys to the fort were always held by the havaldar, not the sarnobat — separating physical control of the installation from military command.1
This was not bureaucratic convention. It was an anti-corruption and anti-treachery design. If a Mughal agent wanted to suborn a fort, he had to successfully compromise all three officers simultaneously — a dramatically harder target than bribing a single commander. The deliberate mixed-caste, mixed-community appointment policy made coordinated treachery even harder by ensuring the three officers had no pre-existing loyalty network to exploit.
The havaldar held the keys. This single detail is the load-bearing structure of the whole system: the person with the most intimate physical control of the fort's access points was specifically not the person with military command authority. Authority was separated by design, not by accident.
Rajgad's Three-Machi System. Rajgad (Shivaji's early capital) was engineered with three separate fortified plateaus (machis) connecting to a central citadel — creating a defensive architecture in which capturing any one section did not give access to the others. An attacking force that breached the outer machi found itself in an enclosed killing ground with the inner machis still defended. The fort could be partially taken and still function. Redundancy was built into the stone.1
Purandar at 4,500 Feet. Purandar's elevation made conventional siege approaches impossible for artillery heavy enough to breach its walls. The Treaty of Purandar (1665) — in which Shivaji ceded 23 forts to the Mughals — was not a sign that fort architecture had failed; it was a sign that the sheer number of forts made the loss of any 23 of them survivable. The network was deliberately over-built to absorb setbacks.1
Raigad as Capital: Strategic Site Selection. When Shivaji chose Raigad as his capital and coronation site, the selection criteria were explicit: sea access for supply and escape routes combined with near-impregnability from land assault. The fort sat atop a plateau with a single narrow approach path that could be held against any army. A capital is typically chosen for symbolic or commercial reasons; Raigad was chosen for survival probability. This inverts the conventional logic of capital city selection — commerce and symbol follow military logic, not the other way around.1
After the Treaty of Purandar (1665) surrendered 23 forts to Mughal authority, Shivaji spent three years in apparent quiescence and then, in 1670, recaptured 23 forts in approximately four months. The speed of recapture was only possible because the Maratha administrative and military network had been maintained in the surrounding territory during the Mughal occupation. The forts were isolated installations the Mughals held but could not use as nodes in an integrated administrative network — they were military outposts, not political units, while in Mughal hands.1
This revealed the asymmetry in what the same physical structure meant to each side. To the Mughal empire, a Deccan hill fort was a garrison post. To the Maratha state, it was the node of an information, supply, and administrative network. The same stone had different strategic value depending on whether the occupier could activate its political function.
[POPULAR SOURCE] — Purandare describes fort architecture and administration in detail but without citing engineering records or administrative documents. The havaldar/sabnis/sarnobat triad and the key-holding convention are described as standard Maratha practice across all forts, which is plausible given the consistent pattern but requires corroboration from administrative records.1
Tension with the concentration-of-force principle: Sun Tzu and Clausewitz both argue for concentration of force at the decisive point rather than distribution. Shivaji's 60+ fort network was the opposite — deliberate distribution of military capacity across many points. The distributed system traded striking power for resilience: no single defeat could collapse it. Whether this trade-off was consciously made or was a product of the specific geographic and political constraints is unclear.
History — Sun Tzu Nine Situations: Sun Tzu — Nine Situations and Desperate Ground — Sun Tzu's nine ground types (from dispersive through desperate) are organized by retreat cost. Shivaji's fort network systematically managed retreat cost: every major operation was staged from a fort close enough to serve as a fallback. The Panhala–Vishalgad–Raigad sequence during the 1660 siege shows this explicitly — Baji Prabhu's stand at Ghodkhind pass was not heroic sacrifice; it was the tactical rear-guard that made the Panhala-to-Vishalgad transit possible. The fort network provided the "fatal ground" option that Sun Tzu identifies as the source of maximum cohesion — but deliberately structured so the fatal moment came only in the pass, not in the fort itself.
History — Arthashastra Kingship and the Rajarshi Ideal: Arthashastra — Kingship and the Rajarshi Ideal — Kautilya's exhausting royal schedule functions as structural anti-corruption: the king is always present, always checking, which makes it harder for any subordinate to accumulate corrupt power. The havaldar/sabnis/sarnobat triad is an architectural version of the same principle — instead of the king's presence preventing power consolidation, the design of the institution prevents it. Both address the same problem (how does a ruler prevent subordinates from accumulating loyalty that could turn against the ruler?) with different tools. Kautilya's answer is personal surveillance; Shivaji's answer is structural separation of authorities. The comparison produces an insight: personal surveillance scales poorly (the ruler can only be in one place); structural separation of authorities scales across 60+ forts.
The Sharpest Implication The fort system is a theory of resilience that most organizational designers miss: resilience is not built by making the central node stronger — it is built by making the failure of any single node irrelevant to the functioning of the whole. The Maratha state could lose Purandar and survive. It could lose 23 forts at once and rebuild in four months. This is only possible when the network is designed so that no single point is load-bearing. The implication for any distributed system — military, organizational, informational — is that resilience is an architectural property, not a strength property. Making the central command stronger does not make the system more resilient; distributing capacity and making nodes separable does.
Generative Questions
The architecture appears as named places. Rajgad the First Capital gives the three-machi redundancy its own treatment as the first capital's design. The Siege of Purandar 1665 is the half-way-up artillery doctrine that finally cracked the hill-fort design — the case where the engineering met an attacker who solved it. And Raigad the Capital Fortress documents the survival-first site selection that made the coronation seat near-impregnable from land.