Infrastructure for long-running agents

Agents thatkeep running.

Q: Is Molted another VPS or Kubernetes wrapper?

No. A generic cloud gives you machines and an orchestration primitive. Molted gives you an agent-ready operating environment: runtimes, recovery, versioned state, browser automation, integrations, observability and the control plane around them.

Q: When should I use Molted instead of a sandbox platform?

Use a sandbox when you need an isolated execution environment for a task or session. Choose Molted when you want OpenClaw or Hermes operated as an ongoing service: a persistent workspace and identity, runtime-aware health monitoring and recovery, managed tools and a fleet control plane. Sandbox platforms can also support persistence; Molted's focus is managing the agent's operational lifecycle for you.

Q: Which agent runtimes are supported?

OpenClaw and Hermes Agent are supported today through versioned runtime adapters. Each keeps its native state and behavior while Molted supplies the shared lifecycle, health monitoring, recovery and fleet control around it. Additional runtimes can be packaged on the same contract.

Q: Can we keep our own brand and customer experience?

Yes. Your product remains the customer-facing layer. Molted sits underneath it through the API and white-label domains, or provides tenant dashboards for your team. molted.cloud is built exactly this way.

Q: Can it run on our infrastructure?

Yes. The same platform can run on Molted-managed clusters, on your own infrastructure, or on a dedicated sovereign cluster.

Q: How does Molted recover a failed agent?

Molted monitors the runtime from inside and outside the instance, captures a post-mortem, then escalates through four recovery tiers until the agent is healthy again.

Q: What happens to an agent's files?

Every workspace is versioned automatically. You can inspect a diff, restore a known-good version and recover deleted files without bolting on a separate backup product.

Q: How do integrations and browser automation work?

Agents can use managed browser sessions and connect to more than 1,000 business applications. Credentials remain managed outside the runtime and are never baked into an instance image.

Q: How is Molted priced?

Managed deployments are billed per instance, per day and pro-rated. The price depends on instance count, region, resource priority and commitment, which is why the cost calculator exists: it returns your exact monthly figure. On-premise and Swiss deployments use the same operating layer with commercial terms sized to the fleet.

Q: Are we locked into Molted?

No. Agent state remains plain, portable files; runtimes are version-pinned; and on-premise deployment keeps the full stack on infrastructure you control.

A persistent home for OpenClaw and Hermes. Your agents keep their memory, tools and identity. We handle the infrastructure and recovery, on our cloud or yours.

Talk to our team Estimate your fleet

Managed cloud On-premise White-label

The ongoing life of an agentInteractive illustration

moltedcontrol plane · eu1persistent agent fleet

7F2A open portal.customer.combrowser · managed session

12 renewal records extracted2.4s

X1N8 write memory/findings.md+41 lines · rev 185

Q8C4 reply · re: renewal termsmanaged mailbox

deliveredthread continues

Z1H7 call +41 22 555 01 84dedicated number

fleet healthy · 6/6 heartbeatseu1 · 4s

7F2ARUNNING· 41DRenewals · portal.customer.com

Node: eu1-07
Workspace: rev 184
Identity: preserved

OPERATION 01 / 07

7F2A open portal.customer.com

browser · managed session

~900

peak concurrent agents

99.9%

uptime track record

300+

managed clients

<18s

to provision

Our conviction

Making agentsas obvious as chatbots.

Deploying a fleet of autonomous agents should feel as simple as shipping a chatbot. Molted makes that possible.

You shape the experience. We run everything that keeps it alive.

01Built for continuity

The next task.
The same agent.

A session sandbox is a natural fit for an isolated job: execute code, return a result, finish. An autonomous agent has an ongoing life, across requests, conversations and days of work.

Molted is built around that continuity. Keep the agent's workspace and identity, give it managed tools, and recover its runtime when something fails. Operate a whole customer fleet from one control plane.

One agent. An ongoing workspace.Illustrative lifecycle

Deploy
OpenClaw or Hermes, ready to work.
Pick up the next task
New request. Existing context.
Recover the runtime
Recreate compute. Reattach the workspace.
Keep operating
The same agent, serving your customer.

Persistent across the lifecycle

Agent identity
Versioned workspace
Managed tool access

02Agent runtimes

Your runtime.
Our responsibility.

OpenClaw and Hermes are ready today. Need another agent or a custom harness, including Claude Code? We package it around the same operating contract: release, workspace, health and recovery, while preserving its native experience.

OPEN-SOURCE PERSONAL AGENT

OpenClaw

Live

Multi-channel agents with skills, scheduled work and persistent conversations, running in a dedicated instance.

Workspace: Persistent
Recovery: Managed
Release: Version-pinned

AGENT BY NOUS RESEARCH

Hermes

Live

Hermes Agent with its native tools, conversations and dashboard, running in a dedicated, supervised instance.

Workspace: Persistent
Recovery: Managed
Release: Version-pinned

CUSTOM HARNESS

Your own runtime

On request

Bring a Claude Code or Codex harness, or another agent runtime. We package its native experience around the same operating contract.

Harness: Yours
Contract: Shared
Packaging: Managed

MoltedOSOne operating contract around every runtime

01Pin release

02Sync native state

03Watch health

04Repair or recreate

OpenClaw + Hermes ready today · Claude Code, Codex and custom harnesses available on request

03Performance benchmark

Hardware selected
for the agent.

We benchmark the runtime itself, then select the host that gets an agent ready fastest. The result is capacity tuned for real agent work, rather than a generic cloud SKU.

Runtime performance lab

OpenClaw cold-start trial

Median · repeated trials

Mac miniHardware price · $799

5.1sDesktop baseline

MacBook Pro M3 ProHardware price · $3,400

4.0sHigh-end Mac baseline

Molted premiumRuntime-selected hardware

1.1sBest measured result

3.6×

faster than the high-end Mac benchmark.Median of repeated internal trials · OpenClaw cold start · representative hardware

After the benchmark, the operating layer keeps the advantage useful.

Node selectionRuntime-tested host

Fleet capacityLive headroom control

Instance isolationDedicated resource envelope

04The Molted stack

Compute is replaceable.
The agent is not.

Explore what keeps an agent in production. Five responsibilities, one operating layer. Select a layer to see how it works.

01Runtime

Every agent runs as a pinned, operated release.

Choose OpenClaw or Hermes and pin a release. Molted prepares the instance and watches the runtime's health throughout its life.

ReleaseOpenClaw or Hermes, version-pinned

Healthprocess, gateway and heartbeat

Upgradeexplicit, per instance, reversible

01 / 05 · Architecture illustration

05Capabilities

Your agents can reach the work, not just the API.

Managed MCP, browser sessions and more than 1,000 business integrations are part of the environment, not another connector backlog for your team.

Gmail

HubSpot

Slack

Discord

Notion

Salesforce

Instagram

Zoom

Google Calendar

Stripe

Linear

GitHub

Browser automation

Persistent sessions, managed proxies and captcha handling for the web that has no API.

SESSION · PERSISTENT · PROXY · MANAGED

Agent email

Managed mailboxes for sending, receiving and acting on real customer conversations.

IDENTITY · MANAGED · THREAD · CONTINUOUS

Voice and phone

Dedicated numbers, calls, SMS and 2FA workflows when the job crosses channels.

NUMBER · DEDICATED · RECORDING · POLICY CONTROLLED

Browse all integrations

06Built to recover

Recovery happens before the ticket.

Each instance is monitored from inside the runtime and outside the pod. When something fails, Molted captures the evidence, escalates through recovery tiers and restores service automatically.

00:00Process stops respondingobserved

00:05Heartbeat misseddetected

00:12Post-mortem captureddiagnosed

00:47Instance healthyrecovered

Illustrative recovery trace

47sec

resolved

monitoring layers

recovery tiers

Auto

post-mortem capture

24/7

fleet monitoring

Every failed recovery produces a post-mortem and escalates with its diagnostic context intact.

Recovery sequence

Four tiers, always armed.

In-runtime repair

Restart with bounded backoff, then run the runtime doctor when a configuration can be repaired in place.

Clean instance recreation

Rebuild compute and networking on a healthy node while preserving the versioned workspace.

Known-good restore

Roll the configuration back atomically to the last healthy filesystem revision.

Human escalation

Escalate with the post-mortem, health history and diagnostic logs already attached.

State you can inspect and restore

Every file change leaves a trace.

Inspect what changed. Compare revisions. Restore a known-good workspace, even after a delete. Plain, portable files keep the agent's state independent of its compute.

AutomaticEvery workspace revision is captured

InspectableDiff files between any two points

PortablePlain files, no proprietary memory format

workspace / atlas-prod

saved

Files

▾ workspace

▾ memory/

findings.md

▾ skills/

customer-research.md

SOUL.md

rev 184 · 12s ago

customer-research.mdv183 → v184

# Customer research skill

- Summarize the first result

+ Cross-check primary sources

+ Flag conflicting evidence

Return findings with citations

Illustrative workspace diff · plain files, versioned

07Your product on top

Your brand in front. Molted underneath.

Own the experience your customers see. Molted runs the agents underneath, through an API and white-label domains. We built our own product on that same foundation.

We are our first customer

A real product.
The same foundation.

molted.cloud is our customer-facing agent product: its own experience, customers and billing. Its agents run on the same Molted infrastructure available to you.

Explore molted.cloud

Public molted.cloud demo showing an agent completing an invoice follow-up mission — molted.cloud · public product demonstration · captured September 2026

01 · Your product

The customer relationship

Your interface, brand, workflows and billing.

02 · The connection

API + white-label domains

Provision and manage agents from your own product.

03 · Molted underneath

The ongoing operations

Runtime, versioned state, tools, monitoring and recovery.

Platform surface

The API behind your product.

api.eu1.molted.net

request example

Create instance

POST /v1/instances
{
  "tenantId": "YOUR_TENANT_ID",
  "handle": "support-agent",
  "runtime": "openclaw",
  "resourcePriority": 7
}

RuntimeOpenClaw

StateVersioned

NetworkIsolated

RecoveryEnabled

managed instance

Complete lifecycle API

Create, configure, start, stop, inspect and recover every instance.

White-label by design

Custom domains, tenant dashboards and your brand in front of the platform.

Independence by design

Any model provider, open runtimes and plain versioned files. Leaving is a transfer, not a rewrite.

Authentication for every surface

API keys for your backend, scoped instance tokens for end-user frontends and user accounts for the team dashboard.

Read the API docs

Lifecycle surface

The operations your product needs, without a second control plane.

Region-scoped endpoints, tenant-aware authorization and stable instance identifiers across restarts.

POST/v1/instances

Provision an instance

GET/v1/instances

List and filter a fleet

PATCH/v1/instances/:id

Change desired state

POST/v1/instances/:id/restart

Recover or redeploy

GET/v1/instances/:id/logs

Stream runtime logs

GET/v1/instances/:id/events

Read the event timeline

POST/v1/instances/:id/commands

Execute an operation

GET/v1/instances/:id/fs/tree

Inspect workspace state

DELETE/v1/instances/:id

Soft-delete safely

API keys for backendsScoped instance tokensSSE log streamingLifecycle event timeline

08Deployment & pricing

Operated where your business requires.

Start on our infrastructure, bring your own, or isolate the entire stack. Every instance gets isolated compute, storage, network policy and identity. The operating model stays the same.

One operating model

The same stack.
On your terms.

Runtime, state and control plane.
Deployed together, wherever you run.

Managed cloud

Molted-operated regions

We run the platform and the infrastructure. Choose your region and start with no platform team to build.

Regional hosting · per-instance, per-day billing

99.5% infrastructure SLA10% MRR credit if breached

On-premise

Your infrastructure boundary

The full stack runs on your machines, including the control plane. Your network, your hardware, your data.

Air-gapped deployment available · no setup fee

99.9% infrastructure SLA15% MRR credit if breached

Swiss cluster

Dedicated infrastructure in Switzerland

A dedicated, isolated deployment for workloads with stricter residency requirements. Operated by Molted.

Swiss data residency · managed operations

99.9% infrastructure SLA15% MRR credit if breached

Size your fleet. See your price.

Per-instance, per-day billing. Choose the region, priority and commitment in the calculator to get the price for your fleet.

Open cost calculator

The SLA covers infrastructure availability: instance, network and API. Agent application logic is excluded.

09Build vs operate

Do not become an infrastructure company by accident.

The real cost is not the node. It is the platform team, on-call, integrations and months spent building everything around it.

Compare your current stack

Solutions by team

RequirementDIY cloudMolted

Runtime setupYour teamManaged

24/7 incident responseYour teamIncluded

Workspace versioningBuild itIncluded

Business integrationsBuild each1,000+

Fleet capacityPlan and monitorManaged scheduling

Recovery systemBuild and maintain4 tiers

White-label surfaceBuild itDomains, API, dashboards

10Questions

The short version.

Is Molted another VPS or Kubernetes wrapper?

No. A generic cloud gives you machines and an orchestration primitive. Molted gives you an agent-ready operating environment: runtimes, recovery, versioned state, browser automation, integrations, observability and the control plane around them.

When should I use Molted instead of a sandbox platform?

Use a sandbox when you need an isolated execution environment for a task or session. Choose Molted when you want OpenClaw or Hermes operated as an ongoing service: a persistent workspace and identity, runtime-aware health monitoring and recovery, managed tools and a fleet control plane. Sandbox platforms can also support persistence; Molted's focus is managing the agent's operational lifecycle for you.

Which agent runtimes are supported?

OpenClaw and Hermes Agent are supported today through versioned runtime adapters. Each keeps its native state and behavior while Molted supplies the shared lifecycle, health monitoring, recovery and fleet control around it. Additional runtimes can be packaged on the same contract.

Can we keep our own brand and customer experience?

Yes. Your product remains the customer-facing layer. Molted sits underneath it through the API and white-label domains, or provides tenant dashboards for your team. molted.cloud is built exactly this way.

Can it run on our infrastructure?

Yes. The same platform can run on Molted-managed clusters, on your own infrastructure, or on a dedicated sovereign cluster.

How does Molted recover a failed agent?

Molted monitors the runtime from inside and outside the instance, captures a post-mortem, then escalates through four recovery tiers until the agent is healthy again.

What happens to an agent's files?

Every workspace is versioned automatically. You can inspect a diff, restore a known-good version and recover deleted files without bolting on a separate backup product.

How do integrations and browser automation work?

Agents can use managed browser sessions and connect to more than 1,000 business applications. Credentials remain managed outside the runtime and are never baked into an instance image.

How is Molted priced?

Managed deployments are billed per instance, per day and pro-rated. The price depends on instance count, region, resource priority and commitment, which is why the cost calculator exists: it returns your exact monthly figure. On-premise and Swiss deployments use the same operating layer with commercial terms sized to the fleet.

Are we locked into Molted?

No. Agent state remains plain, portable files; runtimes are version-pinned; and on-premise deployment keeps the full stack on infrastructure you control.

What can we control through the API?

The regional REST API covers the complete instance lifecycle: provision and update instances, start or stop them, execute operational commands, stream logs, inspect event history and workspace state, and recover or delete an instance. API keys, scoped instance tokens and user sessions serve different product surfaces.

How are model and integration credentials protected?

Credentials are encrypted at rest and resolved through controlled platform services. They are not baked into runtime images or shared between tenants, and instance capabilities receive only the access required for that instance and request.

production infrastructure, already operated

Build the agent product. We will run what is underneath.

Start with a technical walkthrough or a 50-instance pilot, with free setup, on your infrastructure or ours.

Talk to our team Read the API docs

Agents thatkeep running.

The next task.The same agent.

Deploy

Pick up the next task

Recover the runtime

Keep operating

Your runtime.Our responsibility.

Hardware selectedfor the agent.

OpenClaw cold-start trial

Compute is replaceable.The agent is not.

Every agent runs as a pinned, operated release.

Your agents can reach the work, not just the API.

Browser automation

Agent email

Voice and phone

Recovery happens before the ticket.

Four tiers, always armed.

In-runtime repair

Clean instance recreation

Known-good restore

Human escalation

Every file change leaves a trace.

Your brand in front. Molted underneath.

A real product.The same foundation.

The customer relationship

API + white-label domains

The ongoing operations

The API behind your product.

Complete lifecycle API

White-label by design

Independence by design

The operations your product needs, without a second control plane.

Operated where your business requires.

The same stack.On your terms.

Managed cloudManaged cloud

On-premiseOn-premise

Swiss clusterSwiss cluster

Do not become an infrastructure company by accident.

The short version.

Is Molted another VPS or Kubernetes wrapper?

When should I use Molted instead of a sandbox platform?

Which agent runtimes are supported?

Can we keep our own brand and customer experience?

Can it run on our infrastructure?

How does Molted recover a failed agent?

What happens to an agent's files?

How do integrations and browser automation work?

How is Molted priced?

Are we locked into Molted?

What can we control through the API?

How are model and integration credentials protected?

Build the agent product. We will run what is underneath.

The next task.
The same agent.

Your runtime.
Our responsibility.

Hardware selected
for the agent.

Compute is replaceable.
The agent is not.

A real product.
The same foundation.

The same stack.
On your terms.

Managed cloud

On-premise

Swiss cluster