Two sovereign chains cannot talk to each other on their own. IBC defines how they should, but the protocol only describes the handshake — it does not move the packets. Vitwit runs the relayers so that when something stalls, the someone who notices is us, before you ever find out.
IBC lets independent chains transfer tokens and arbitrary messages while each keeps its own validator set. A relayer is the off-chain process that carries packets between them.
The difficulty is that a relayer is a long-lived process babysitting infrastructure it does not control. Chains upgrade on their own schedules, breaking relaying silently. Packets have timeouts — miss the window and the transfer fails. Channels expire if a client isn't kept fresh. Gas has to be funded on every chain the relayer touches. And the failure mode is quiet: nothing crashes, packets just stop flowing, and the first signal is often a user asking where their funds went.
Running one relayer for a weekend is easy. Running dozens across mainnet, through every upgrade, for years, is an operational discipline.
The point is that packets keep moving across every path you depend on, through every upgrade, without your team owning the pager.
Not a test setup. Operating at fleet scale means we've already seen the edge cases a new path will hit, and the tooling and monitoring are built for dozens of connections running simultaneously.
We hold service-level objectives on channel health, not just 'the process is running.' Packet flow, acknowledgement latency, and stuck-transfer detection are measured and alerted on — a degrading channel is caught while it's degrading, not after funds are stuck.
Chain upgrades are the single most common way relaying breaks. We've stayed healthy through every upgrade since IBC launched. Recovery is automated and the team has run the playbook enough times that an upgrade is a routine event, not an incident.
Need a new channel to a chain you're integrating? Onboarding typically takes days, not weeks. The infrastructure to stand up, fund, and monitor a new relayer is already in place — adding a path is configuration, not a project.
Relaying rarely stands alone. A chain that needs IBC paths usually also needs validators, RPC endpoints, and often launch and upgrade engineering. Fragmenting that across vendors means no one owns the seams — when a chain upgrade breaks both the validator and the relayer at once, you do not want two vendors debating whose problem it is.
Because we contribute to the IBC stack itself and operate validators on the same chains we relay between, an upgrade that affects relaying is one our team has usually already prepared for. One SLA, one contract, one on-call rotation covering the whole path a packet travels.
Talk to the infra team →The cost of a stalled channel is measured in stuck user funds and support load. We back it with a 99.99% uptime floor, HSM-backed key custody, active-active deployment across AP, EU, and US, and a 24/7 NOC. The team that has relayed through every IBC upgrade without losing a path is the team that picks up when you call.