Autonomous infrastructure agent

Your L2 protocol
has a new
operator.

Nocca watches your nodes, runs your upgrades, scales your infra, and pages you only when it genuinely can't handle it. The rest of the time? It works while you sleep.

nocca --protocol=base-mainnet
06:14:02 NODE base-sequencer healthy (latency 4ms)
06:14:02 NODE base-full-node-03 healthy (blocks: 21,847,291)
03:47:11 UPGRADE v2.4.1 queued — staging rollback ready
03:47:13 UPGRADE deployed to base-sequencer — OK
03:47:18 UPGRADE deployed to base-full-node-03 — OK
03:47:19 UPGRADE all nodes updated in 8s — no human involved
00:00:00 SCHEDULE next upgrade window: 2026-06-02 03:00 UTC
142 upgrades run
0 pages sent
99.97% uptime
The problem

Running L2 infrastructure shouldn't require a human SRE to be awake at 3am.

NodeSmith at Coinbase runs 60+ chains with a "small, specialized team." That's the hint. The team is small because humans are expensive. But the monitoring never stops. The upgrades never stop. The incidents never stop.

The crypto ops stack is full of tools built for developers — not for autonomous operation. Dashboards for humans to interpret. Alarms for humans to respond to. Code for humans to write.

What the industry actually needs is an autonomous operator. One that owns the infrastructure end-to-end. One that pages a human only when it genuinely cannot solve the problem itself.

Not a dashboard.
A worker.

24/7 Node Monitoring

Tracks block height, latency, sync status, and peer connectivity across every chain you're running. Alerts fire only when thresholds are breached, not on every metric change.

Autonomous Upgrades

NodeSmith cut Coinbase's upgrade effort by 30%. Nocca does the same for every L2 — tests the upgrade in staging, rolls forward with a rollback plan ready, and reports the outcome. No human in the loop.

Incident Triage

When something breaks, Nocca runs the playbook: collect context, identify the likely cause, apply the fix, verify recovery. Pages you with a full incident report only if the fix didn't work.

Scaling Orchestration

Detects capacity pressure and spins up or down node replicas based on real load, not schedule. Scales with your protocol — you define the rules, Nocca enforces them.

Staged Rollback

Every change Nocca makes includes a rollback path. Upgrade fails in staging, you never see it. Upgrade fails in production, it reverts in under 60 seconds. Human oversight is guaranteed — but human involvement is optional.

Daily Digest

Every morning: a plain-English summary of what ran, what changed, what needs attention. No graphs to interpret. No dashboards to open. Just the state of your infrastructure in 200 words.

Three layers. One agent.

observer

Polling your node RPCs, block explorers, and cloud monitoring APIs every 30 seconds. Building a picture of current state vs. desired state.

reasoner

LLM-powered decision engine. Given a state delta, it picks the action from its playbook — upgrade, scale, restart, escalate, or wait. All reasoning is logged and auditable.

executor

Takes the reasoned action and runs it against your infrastructure — via RPC calls, cloud provider APIs, or SSH. Applies the change, verifies it landed, logs the outcome.

Every action has a staged rollback. Every decision links to the state that triggered it. If Nocca can't resolve something, you get a page with full context — not a vague alert.

Protocols deserve operators
that never need to sleep.

Nocca was built for the infrastructure engineers who've done the 3am pager shift. It doesn't replace you — it replaces the part of your job that shouldn't require a human.

Protocols supported Base, Arbitrum, Optimism, zkSync, Starknet, Solana
Recovery SLA < 60 seconds from failure to rollback
Onboarding Connected and monitoring in under 15 minutes
Escalation Plain English page with full context — never raw alerts

The SRE rotation is a human invention. Nocca makes it optional.