Summary: MACI is best understood not as a voting app, but as privacy-preserving tally-and-message infrastructure for onchain voting and funding rounds. Its core mechanism is a split stack: voters publish encrypted messages onchain, a coordinator manages poll setup and tallying, zk proofs certify correct execution, and optional relayer / coordinator services make the system usable by downstream apps. That makes MACI a strong comparison class for Vocdoni, Snapshot-adjacent governance stacks, clr.fund, and DAO voting plugins. The real control surface is not only who can vote, but who runs the coordinator, who defines gatekeeping and voice-credit rules, how much trust remains in tally decryption, and whether privacy and bribery-resistance live at the protocol layer or get re-centralized in operational services.
What it does:
Provides an Ethereum-based protocol for private, collusion-resistant voting where individual votes stay hidden but final tallies remain publicly verifiable
Uses encrypted vote messages plus zk-SNARK proofs so vote processing and tally publication can be audited without revealing each voter’s choice
Supports both quadratic and non-quadratic voting / funding flows, which is why projects like clr.fund and Gitcoin-style funding rounds built on it
Exposes a growing middleware surface around the core protocol, including SDK, relayer, coordinator service, subgraph, and packages for contracts, circuits, cryptography, and state-machine logic
Adds poll-level customization such as custom gatekeeping and custom voice credits, making voter admission and voting power configurable rather than fixed by one baked-in identity system
Is being pushed toward DAO-governance integrations through plugins and tooling-provider integrations rather than remaining only a standalone public-goods-funding primitive
Key claims:
MACI’s main reusable primitive is not “private voting” in the abstract, but receipt-free onchain voting with a trusted coordinator plus zk-verified tally correctness. That coordinator dependency is the most important thing not to flatten away.
The official docs are explicit that collusion resistance, privacy, and receipt-freeness hold only against everyone except the trusted coordinator. A dishonest coordinator cannot forge the final tally if proofs verify, but can still see votes in cleartext and potentially collude with bribers. That is the protocol’s clearest trust bottleneck.
The 2025 roadmap makes the control-plane problem even clearer. MACI is productizing coordinator services, relayers, and DAO integrations while simultaneously researching coordinator decentralization via MPC, homomorphic encryption, or TEEs. In other words, the team itself treats coordinator trust as the central unresolved issue.
MACI v3’s custom voice credits and custom gatekeeping matter analytically because they shift practical power into poll configuration. The protocol may protect vote privacy, but organizers still shape who counts and how much each voter can spend.
The repo structure shows MACI is becoming a full middleware stack: circuits, contracts, SDK, subgraph, relayer, and coordinator packages. That makes it more comparable to governance infrastructure providers than to a single voting app.
The separate MACI Platform repo is useful mainly because it is no longer maintained. That sharpens the distinction between the actively maintained protocol layer and the application shell that communities may need to fork or operate themselves.
MACI belongs in the active corpus because it isolates a governance tradeoff many systems hide: privacy and anti-bribery can improve without eliminating concentrated authority, because tally privacy, poll deployment, relaying, and gatekeeping may still collect around coordinator operators and integration partners.
Whitepaper: No single canonical MACI whitepaper surfaced in this pass. The strongest primary materials were the official docs, roadmap, protocol repository README, and MACI Platform README; see ../whitepapers/maci-primary-sources-2026-05-10.md.
Governance-backend contrast: vocdoni when the main question is whether privacy and tally integrity live in a dedicated election stack or in coordinator-run zk middleware.