Solana-anchored coordinator
Every supported chain shares one finality clock. No per-route governance, no bespoke validator set per corridor — one source of truth running on Solana mainnet.
build passing · 92% test coverage · MIT
One contract.
One contract.
Fifty chains.
Four hundred milliseconds.
Switchboard turns 50+ blockchains into a single state machine. Write once, read everywhere — at trading-grade latency, with the gas math to back it up. Built on a Solana coordinator, delivered as a 2-service stack you can run yourself or consume as a managed API.
Read the source MIT · 92% test coverage
- Coordination latency
- <400ms
- Chains supported
- 50+
- Ops / second
- 10,000+
- Uptime SLA
- 99.9%
$ switchboard route open \
--from ethereum \
--to solana,arbitrum,base,sui \
--payload ./order.json
[12:04:17.812] coord → solana slot 248_117_902 ✓ 0ms
[12:04:17.901] verify → arbitrum light-client ✓ 89ms
[12:04:17.943] verify → base light-client ✓ 131ms
[12:04:18.064] verify → sui bls-quorum 12/15 ✓ 252ms
[12:04:18.182] ack → ethereum receipt 0xc4f… ✓ 370ms
route 0x6a…f31 settled · p50 252ms · p99 370ms 5 chains · 1 commit p99 370ms
Architecture
A coordinator, not a committee.
Most cross-chain stacks invent governance per route. Switchboard collapses the problem onto a single Solana program that every chain agrees to watch.
Sub-400ms end-to-end
From source commit to destination verification, the median route lands inside 400ms. That is bridge throughput at orderbook speed.
50+ chains, one SDK
EVM L1s and L2s, Solana, NEAR, Cosmos, Sui, Aptos, plus emerging chains like Celestia and StarkNet. One TypeScript SDK, one HTTP API.
2-service architecture
A Customer API gateway (auth, rate limiting, proxy) and a Core Engine (oracle, chain ops, billing, coordination) — nothing else. MongoDB or PostgreSQL behind it. You can read the whole topology on a napkin.
Pluggable verifiers
Each chain uses the strongest available proof: light-client receipt where possible, BLS threshold signature otherwise, zk-proof for chains that ship them.
40% gas savings via batching
The coordinator commits messages in batches, so destination chains amortize calldata costs across the bundle instead of paying per-route.
Bring-your-own signers
Enterprise teams can plug their own threshold-signing nodes into the relayer pool, keeping signing material under their own custody.
2-service stack
You could draw it on a napkin.
Customer API handles auth, API gateway, rate limiting, security, request proxy. Core Engine runs the oracle service, blockchain ops, billing logic, and cross-chain coordination. That is the whole topology. MongoDB or PostgreSQL behind it — pick with one env var.
- 01Auth + proxy layer (port 3000)
- 02Oracle + chain ops (port 3001)
- 03Pluggable persistence
- 04Solana coordinator program
┌──────────────────┐ ┌──────────────────┐
client ──▶│ Customer API │────────▶│ Core Engine │
│ :3000 │ │ :3001 │
│ auth · rate │ │ oracle · chain │
│ proxy · keys │ │ billing · coord │
└──────────────────┘ └──────────────────┘
│
▼
┌──────────────────────────────┐
│ Solana coordinator program │
│ state-oracle + coordinator │
└──────────────────────────────┘
│
┌──────────────────────┼──────────────────────┐
▼ ▼ ▼
ethereum arbitrum sui
light-client light-client bls-quorum Honest comparisons
How Switchboard stacks up.
We do not pretend competitors do not exist. We picked a position on the curve — speed and simplicity — and these pages are the receipts.
vs read →
LayerZero
A messaging fabric with chain-paired DVN/Executor security. Wide chain support, configurable security stack.
vs read →
Wormhole
A 19-guardian message-passing network with a strong ecosystem footprint and battle-tested ops.
vs read →
Axelar
A Cosmos-style PoS network of validators acting as a general message-passing layer.
vs read →
Chainlink CCIP
Chainlink Labs message-passing with strong institutional governance and risk management framework.
vs read →
Hyperlane
Permissionless interchain messaging with modular ISMs (Interchain Security Modules).
FAQ
The questions we get.
Mostly from infra teams about to wire us into a perp DEX or a real-time game state engine.
How does Switchboard hit sub-400ms across so many chains?
+
Source events get committed to the Solana coordinator in the next slot (~400ms). The Core Engine watches the coordinator, packages the destination payload, and a competing relayer pool races to land it on the destination — usually within the same wall-clock budget.
Is the core actually open source?
+
Yes — MIT-licensed. The monorepo contains the Solana programs, the Customer API, the Core Engine, the SDK, and a demo app. The managed service adds relayer ops, signing infrastructure, and observability.
What happens if Solana halts or has a slow slot?
+
Each route declares a finality policy: fail-open (use the last confirmed coordinator state and continue), fail-closed (pause until the coordinator catches up), or fallback (switch to a secondary verifier). Defaults are conservative.
Can I self-host Switchboard?
+
Yes. `npm run docker:up` brings up both services with Postgres or Mongo. You still need to run relayers and configure signers, which is what most teams end up paying us for.
How is message security enforced?
+
Switchboard ships a verifier per destination: light-client proof where the chain supports it, BLS threshold signature from the relayer pool otherwise, and zk-proof receipts on chains that natively verify them. You can mix and match per route.
What about chains you do not list?
+
New chains land roughly monthly. If you have a specific chain blocking adoption, Enterprise customers can sponsor a chain integration with a 6-week ETA.
Does Switchboard handle arbitrary call data, or only token transfers?
+
Arbitrary opaque payloads. Adapters handle on-chain encoding. The protocol does not assume a token model.
What database does Switchboard use?
+
Either MongoDB or PostgreSQL — both are first-class. MongoDB is the default in docker-compose; switch with DATABASE_TYPE=postgresql. The Core Engine abstracts the persistence layer so you pick whichever your ops team already runs.
Wire it up in an afternoon.
git clone, npm run docker:up, and the Customer API is live on :3000 with the Core Engine on :3001. MongoDB by default, PostgreSQL with one env var. Open-source MIT, 92% test coverage, no signup wall.