Managed Compute
Start with capacity provisioned and operated by IDYL.
The Open Compute Network
The world’s compute.One platform.
Connect compute from anywhere. Organise it into governed Subnets.
Make it available through one composable, programmable platform.
Start with a Workspace
Compute included to get you started.
01 Connect compute from anywhere
02 Organise it into governed Subnets
03 Build through one platform
Build on IDYL
From connected compute
to complete applications.
Run code, serve models, build Agents, and coordinate work through one programmable platform.
Any source. One platform.Read the docs
Bring your compute together.
Use the compute that suits your work. Compose capacity across clouds, private infrastructure, and providers through one programmable platform. Your Subnets govern which capacity can participate and where Workloads run.
Customer Cloud
Use capacity in your own cloud account.
Private Compute
Connect machines, servers, and datacentres you own or operate.
Provider Compute
Access capacity supplied by approved providers on IDYL.
Compute governed by youExplore governance
Your compute. Your rules.
Compose a Subnet around your requirements. Choose the machines, providers, and regions that can participate, who can deploy, and the access, runtime, and isolation policies their Workloads must follow.
Build around data residency, private infrastructure, or the safeguards and agreements needed for HIPAA. The operator defines the environment and its operating requirements.
Your capacity mixYour participation rulesYour operating boundary
Execution for every kind of workRead the docs
Put your code to work.
Run applications with Pods and Deployments, complete batch work with Jobs, or open isolated Sandboxes. Place each Workload on suitable capacity under your Subnet’s rules, and follow its status, logs, and results through one execution model.
Pods & DeploymentsJobs & SandboxesGoverned placement
Connectivity for your applicationsDiscuss your use case
Connect the work you run.
Connect Workloads through Services, expose web applications, and control which traffic can reach them. Define private connectivity, public access, and outbound policy alongside your execution resources, so communication follows the requirements of your environment.
Service connectivityWeb application accessTraffic policy
Your durable computerExplore a Workspace
Keep everything you build.
Keep files, context, environment configuration, outputs, and Run history in a Workspace. Attach execution when there’s work to do, and return to the same saved state when you need it again.
Persistent filesRun historyOn-demand execution
Model execution on your termsDiscuss your use case
Put models to work.
Serve open-weight models on suitable CPU or GPU compute. Call Inference directly, or bring model execution into your applications, Workspaces, Agents, and Flows through the same placement, access, and policy foundations.
Open-weight modelsSuitable CPU / GPU capacityShared platform policy
Durable software actorsDiscuss your use case
Give software the ability to act.
Build Agents with durable identity, private memory, and an assigned Workspace. Connect model access, tools, and Channels through scoped capabilities. Define and version their behaviour, and retain context as individual executions start and stop.
Identity & memoryScoped capabilitiesDurable working context
Communication with continuityDiscuss your use case
Keep people and software connected.
Bring Agents, people, Flows, and applications into governed Channels. Retain messages, membership, delivery state, and history while participants are idle. Carry approvals and decisions forward to activate Agents or signal Flows when participants are ready to act.
Governed membershipDurable deliveryRetained history
Durable coordinationDiscuss your use case
Coordinate work from start to finish.
Connect applications, Agents, people, and external events in a durable process. Fan out work, wait for a decision, retry a step, and resume when an event arrives. Pause without holding compute for the wait. Keep the state and lineage across the whole process.
Waits & retriesParallel executionEnd-to-end lineage
Use each resource directly. Compose them through the same identity, policy, state, and execution model.
Explore the docsYour first five minutes
Start with a Workspace.
Keep everything you build.
Bring your code. Run useful work. Keep the files, context, and results in one durable computer.
Create your Workspace Starter compute included. No credit card required.Inside a Workspace
Interactive demo
Files
The code, inputs, and results of your work.
results/frontier.svgPreview
Synthetic design study
Saved resultFind the strongest designs.
256candidates explored
18on the shortlist
Candidate designsBest trade-offs
Shell
Interactive sessions in this Workspace.
Runs
Recorded work, results, and retained history.
OutputExit code 0
Workloads
The execution attached to this Workspace.
Artifacts
Outputs kept beyond the execution that produced them.
Storage
Working files and saved results, retained between sessions.
Workspace files
Environment
The configuration this example uses to run.
- Runtime
- Python 3
- Dependencies
- Python standard library
- Compute
- IDYL Managed
- Entry point
search.py
History
Recorded work and the outputs it leaves behind.
A completed design search, with its code, inputs, and results intact. Open a Shell or run it again.
From connected capacity to applications
Connect your compute.
Decide how it’s used.
Compose a Subnet around your purpose. Choose its capacity, participants, and operating rules, from a shared research effort to a private environment with specific residency and isolation requirements.
IDYLOne programmable platform
01Nodes & Fleets
Connect capacity
Connect physical or virtual machines as Nodes. Organise the capacity you operate into Fleets.
Your cloud, datacentre, or machines02Subnets
Compose your domain
Choose eligible machines, providers, and regions. Set participation, deployment, access, and isolation rules around your requirements.
Your purpose. Your operating rules.03Workloads
Make compute useful
Run code on suitable capacity through a common execution model. Follow its lifecycle, logs, and results.
One way to run your work04Applications
Compose what’s next
Build with Workspaces, Inference, Agents, Channels, and Flows. Each uses the same execution, state, and policy foundations.
From infrastructure to applicationsA common foundation for
ApplicationsAI workloadsBuilds & testsAutomation
For the people building
Start in the Console.
Build into your workflow.
Open a Workspace for your first Run. Use the CLI and API to make IDYL part of the tools and processes you already use.
Read the documentationIDYL SDKs
One platform.
In code.
Provision compute, deploy applications, work in durable Workspaces, and run AI through one programmable platform.
For teams with deeper requirements
Bring your requirements.
Find your path into IDYL.
Talk with us about workload isolation, access policies, and the infrastructure your team operates.
Talk to our team01
Execution requirements
Discuss the capacity, runtime, and isolation your workloads need.
02
Access and ownership
Map your team’s responsibilities to resource access and policy.
03
Infrastructure strategy
Explore how IDYL fits your cloud accounts, private machines, and operating boundaries.
The Open Compute NetworkDocumentation
What will you
build on IDYL?
Your code. Your next idea.
Start with a Workspace and put it to work.