Show HN: Superserve – Firecracker microVM sandboxes for long-running AI agents

[ new ]Claude Managed Agents on SuperserveIntroducing Claude Managed Agents on Superserve→

The open-source sandbox for long-running agents.

Durable Firecracker VMs that pause indefinitely, resume instantly, and run as long as your agents need.

Bring your harness →

Connect your data →

Integrate your stack →

Get startedView docs

[01]Indefinite sessionsNo 24-hour limit

[02]Durable stateSnapshot · fork · resume

[03]Secure by defaultIsolation · Network · Credentials

[04]Open sourceSelf-hostable

[01] Use cases

Your agents need their own computers

Stateful, secure sandboxes for agents that work alongside humans

Agent runtime

// agent.ts — long-running harnessimport { Sandbox } from "@superserve/sdk"const sbx = await Sandbox.create({ name: "claude-sandbox" })await sbx.commands.run("npm i -g @anthropic-ai/claude-agent-sdk")await sbx.commands.run("claude -p 'refactor src/**/*.ts'")await sbx.pause() // resume later with the same state

[02] Durable environmentsParallel · Flexible

Computers that match how agents actually work

Start

Instantly

from fully customizable templates

Snapshot

Anytime

to checkpoint, fork and rollback

Pause

Indefinitely

to save idle costs

Resume

Automatically

from exactly where you left off

[03] SecurityBy default

Secured for AI workloads

Firecracker isolation

Every agent and tool call runs in its own microVM. No shared kernels or cross-tenant state — stronger isolation than containers or local worktrees.

Credentials broker

Agents should not possess credentials. Our broker sits between your agent and the APIs it calls, eliminating exfiltration risk with brokered access.

Network controls

Control egress traffic to specified URLs and IPs. Every outbound connection is logged, every rule is yours.

SOC 2 Type 2 · HIPAA · GDPR

Security documentation and compliance reports available in the trust center.

View trust center →

[04] Open sourceSelf-hostable

Transparent by design.

superserve-ai/superserve

Star on GitHub→

[05] Programmatic controlSDK · API

Best in class agent experience

›Streaming

Execute code and commands in isolated environments with real-time output streaming.

›Durable filesystem

Read, write, and manage files across sessions. State persists exactly as the agent left it.

›Elastic resource allocation

Let your agents request CPU/memory resources on the fly from within the box.

›Preview URLs

Expose ports from inside a sandbox to a public URL for instant previews.

›MCP

Connect to your sandboxes via the Model Context Protocol from any MCP-compatible agent.

›Docker support

Run Docker containers natively inside the sandbox, with no compromises on the tooling your agents depend on.

[06] Pricing

Start for free, pay as you scale.

No credit card to start. Per-second billing.

Compute · vCPU$0.0504/ hour

Memory · GiB$0.0162/ hour

Storage · GiB$0.000108/ hour

Get startedStart free→No card required

Your first sandbox in seconds.

npm install @superserve/sdk

Get startedView docs

✓No credit card requiredOpen source418✓BYOC available