Initially pitched as Butlers.io — a 'cute' brand for deploying AI agent swarms cheaply. The tech worked, but it actually worked better for compliance-centric enterprises in healthcare, finance, and legal.
They needed something more serious: one dedicated, fully isolated agent per client.
Butlers.io was built to solve a real infrastructure problem: breaking the 1-VPS-per-agent model so you could deploy agent swarms cheaply. The byproduct of doing so however, was something we realized suited compliance-centric organizations. They have this common requirement:
"We can't have any wires getting crossed between clients."
That requirement — complete isolation of memory, credentials, persona, and skills between clients — became the product. Butlers.io became Isolated Agents.
We built over the OpenClaw runtime, moving many of the core features to a central control plane. This control plane orchestrates the swarms of agents that spin up on demand in ephemeral Firecracker microVMs — isolated by hardware, not software. Each client gets their own agent with its own memory, credentials, and persona.
No shared filesystems. No always-on processes. No cross-client data.
A Firecracker microVM wakes only when a client message arrives.
Agent runs with its own memory, secrets, and persona. Hardware-level separation.
VM shuts down the instant the task completes. Zero idle cost. Zero residual state.
Pay only for active compute seconds. Ephemeral microVMs eliminate idle server costs entirely — making it economically viable to deploy one dedicated agent per client at scale.
From 100 clients to 100,000 — the unit economics hold.
Regulated industries can't afford crossed wires. Isolated Agents delivers separation at the infrastructure level — not through policy or configuration.
Each agent's memory is private. No data bleeds between clients, ever.
API keys and secrets are scoped to each agent. No shared access grants.
All conversations stored in Postgres. Encrypted, auditable, queryable.
Firecracker microVMs enforce boundaries at the hypervisor level. One compromised agent cannot reach another.
Any business with a provider-client relationship and sensitive data is a fit. One dedicated agent per client means zero chance of information crossing boundaries.
Each client's portfolio, risk profile, and history stays with their agent. Compliance-ready by design.
Patient-specific agents with isolated health records. HIPAA-ready architecture, not just policy.
Matter-specific agents that never cross-contaminate privileged information between clients.
Agent logic, crons, and heartbeats all in one always-on process
Chat history stored on the agent's own disk
Always billing, even at zero load
Orchestration separated from execution
All chat centralized, encrypted, searchable and auditable
Wake, execute, terminate. Pay per second.
Every regulated industry wants AI agents. None of the existing platforms can pass a compliance review. We can.
AWS Lambda runs on it. We're applying the same isolation model to AI agents — battle-tested infrastructure, novel application.
Startup experience in both healthcare and fintech — we know what compliance actually means on the ground. Not just the architecture, but the audits, the requirements, and what keeps regulated teams up at night.
AI agents are being deployed everywhere — but regulated industries are being left behind. Not because they don't want AI, but because no platform can meet their baseline requirement: complete data separation between clients. That's a massive market sitting on the sidelines.
We are uniquely positioned to unlock it. We've built the infrastructure. We've worked inside healthcare and fintech startups. We understand the audits, the requirements, and the architecture.
Isolated Agents isn't a feature — it's the only platform built from the ground up for compliance-first AI deployment.
isolatedagents.com