agentcomponentbackendruntimeapplicationshared
The agent-native way to build AI solutions.
One agent-native CLI to write the component, assemble the backend, deploy, and ship the application. Drive it yourself, or hand it to your coding agent.
Bring your Python or C++ code, your model, your data. The platform handles serving, GPU scheduling, type checks, deploys, applications, on-prem. No glue code, no rewrite from prototype to production.
Pick a path
Three common shapes of work. Each links to the flow that opens it.
Build a ComponentWrap your code or model as a reusable typed unit with parameters, files, and stream contracts.Assemble a BackendPin component releases, wire streams, and let the type system validate before runtime.Ship a SolutionTake a backend onto a runtime, expose it through an application, monitor it in production.
Explore by pillar
The platform in six pillars, mirrored by the sidebar. Open any pillar to go deep.
AgentDrive the whole platform from your coding agent — agent-native CLI, MCP, and JSON contracts.BackendConnect components into a typed dataflow graph, validated by type inference before it runs.ComponentWrap code or models as reusable, typed units. Build in Python or C++ from worked-out patterns.RuntimeDeploy, lease, monitor, and operate backends across cloud, private, on-prem, and air-gapped.ApplicationBuild user-facing apps on a backend and publish reusable solutions through a public URL.SharedThe primitives every pillar shares — files, types, and secrets.
Exact commands live in the CLI
Every page on this site cross-links into the CLI for flag tables, JSON contracts, and operational output. Or browse the index from inside ppl.
ppl docs tree
ppl docs get flows/quickstart
ppl <command> --help