Can anyone run a prover on Starknet and what resources or costs are required

Checked on December 11, 2025
Disclaimer: Factually can make mistakes. Please verify important information or breaking news. Learn more.

This fact-check may be outdated. Consider refreshing it to get the most current information.

Executive summary

Anyone can run a Starknet prover in principle: Starknet’s roadmap and docs state provers are permissionless and the SN Stack, SHARP and open-source prover code make running provers feasible for community operators [1] [2] [3]. In practice, until recent upgrades, most proofs were produced via the shared SHARP service to amortize costs; the S-two (Stwo) open‑source prover and SHARP integration are intended to lower proving time (from ~24 minutes to under 3 minutes for a 2B L2 gas block in S-two MVP tests) and reduce costs by at least half, enabling consumer‑grade hardware and client‑side proving over time [4] [5] [6].

1. Who can run a prover — the published rules

Starknet documentation and FAQs state that sequencers and provers are meant to be permissionless components of the network: “Anyone will be able to set up a prover and create proofs for the validity of new blocks” and the SN Stack is published so community teams can use the components that power Starknet [1] [2]. StarkWare also committed to open‑sourcing the prover under Apache 2.0 as part of the decentralization roadmap, which removes a legal barrier to running your own prover [3].

2. The practical path today — SHARP and shared proving

Starknet currently uses SHARP, a shared proving service that aggregates many Cairo program executions into a single STARK proof to amortize cost. That means most ecosystem participants rely on SHARP (or managed services like Atlantic) rather than running a bespoke prover to avoid high per‑proof overheads [4] [1]. Documentation notes SHARP “is used to reduce costs and improve efficiency within the network,” and the transition of SHARP to S-two is part of the plan to make proving more affordable [4].

3. Resources required — software, hardware and integrations

The software stack is public: SN Stack components (Papyrus, Cairo VM, provers, verifiers), example prover implementations (Stone, Stwo), and community tooling are published on GitHub and the Starknet docs — so the technical prerequisites are available [2] [7] [8]. Hardware needs vary: recursive STARKs and the SHARP model were explicitly introduced to allow smaller, more available compute instances instead of huge monolithic servers, and Starknet materials highlight that S-two was designed to enable consumer‑device proving (phones/laptops) over time [9] [10]. Exact CPU/RAM/IO targets are not specified in the provided sources; available sources do not mention a concrete minimum machine spec.

4. Costs — economic reality vs. marketing claims

Historically, proving was expensive enough that SHARP was necessary to amortize cost; Starknet’s documentation frames SHARP as the current cost‑reduction mechanism [4] [1]. Starknet’s S-two rollout claims large efficiency gains: the S-two MVP promises reducing proof time for a 2B L2 gas block from 24 minutes to under 3 minutes and “decrease proving costs by at least half,” and later material touts orders‑of‑magnitude throughput improvements and consumer‑grade proving [5] [6]. Those are vendor and protocol claims; third‑party reporting repeats them but does not publish independent benchmark cost tables in the provided set [10] [11]. Therefore, exact dollar costs per proof or per‑block on today’s mainnet are not present in current reporting — not found in current reporting.

5. Decentralization incentives and hidden agendas

StarkWare and the Starknet project have an explicit roadmap to decentralize sequencing and proving; open‑sourcing the prover and publishing the SN Stack advance that goal [3] [2]. But there is a practical transition period: a shared prover (SHARP) and managed services remain core to operations today to keep fees reasonable and latency low [4]. The incentives for StarkWare and collaborators to promote S-two and SHARP include reducing fees and unlocking new use cases (and business opportunities around managed proving services and tooling) — those incentives align with decentralization in public messaging but also preserve a phased migration pattern [2] [4].

6. What to do if you want to run one — an actionable roadmap

Start by cloning and testing the SN Stack and related GitHub projects (Papyrus, provers, Lambdaclass’s starknet_stack examples) to run a local chain and watcher‑prover; experiment with the Stone/Stwo codebases and follow SHARP integration docs [8] [7] [4]. Join the Starknet community channels and follow the decentralization roadmap and FAQ for validator/prover participation rules and staking requirements as the network moves to a multi‑sequencer/validator model [2] [1].

Limitations: sources here provide feature claims, roadmaps and developer guidance but do not publish a granular, independent cost schedule or minimum hardware spec — available sources do not mention those specifics [5] [4].

Want to dive deeper?
Who can operate a prover on Starknet — individuals, validators, or only specialized firms?
What hardware and software are required to run a Starknet prover locally?
How much does it cost (capital and ongoing) to run a Starknet prover in 2025?
What bandwidth, storage, and compute benchmarks are needed for timely Starknet proof generation?
Are there managed prover services or pools for Starknet and how do fees and trust models compare?