# Chains & steps (/pipeline-sdk/chains)



A pipeline is a directed graph of **steps**. You build it by chaining method
calls off a root and returning the leaf (or leaves). A step built off another
runs after it and inherits its filesystem; steps returned together with no
dependency between them run in parallel.

## Starting a chain [#starting-a-chain]

`hm.sh(...)` starts a chain and sets its first command — shorthand for
`hm.scratch().sh(...)`.

<DslSignature name="sh" />

**Python**

```python
import harmont as hm

build = hm.sh("make build").sh("make test")
```

**TypeScript**

```typescript
import { sh } from "@harmont/hm";

const build = sh("make build").sh("make test");
```

## Forking and joining [#forking-and-joining]

Call `.fork()` to branch, and `hm.wait()` for a join barrier.

<DslSignature name="wait" />

**Python**

```python
import harmont as hm

root = hm.sh("make deps")
lint = root.fork("lint").sh("make lint")
test = root.fork("test").sh("make test")
```

**TypeScript**

```typescript
import { sh } from "@harmont/hm";

const root = sh("make deps");
const lint = root.fork({ label: "lint" }).sh("make lint");
const test = root.fork({ label: "test" }).sh("make test");
```

The full `Step` field and method list is in the
[chains reference](/pipeline-sdk/reference/chains).