orchardCode-mode workflow generation and execution

Generated TypeScript workflow artifacts for CLI-harness-backed agents, executed on a small in-process runtime.

First run tutorial

Package reference

Tutorials

Guided first passes — install packages, initialize Orchard config, generate an artifact, and run it in process.

How-to guides

Task-focused recipes for running harness calls, generating workflow artifacts, and inspecting a local run.

Reference

Workspace packages, exports, and CLI artifact tools.

Explanation

How Orchard layers generated workflow code over an in-process task runtime.

Execution is in-process and in-memory

Nothing is persisted

Orchard executes workflows as promises inside the current Node process through createLocalRuntime(). There is no database, no persistence, no crash recovery, no cross-process execution, and no post-run inspection. When the process exits, the run and every memoized step are gone. Treat this as an in-process workflow runtime, and do not rely on it anywhere losing an in-flight run is unacceptable.

Documentation map

Orchard's core loop is code-mode workflow generation and execution:

text

prompt or workflow spec
  → generated TypeScript workflow artifact
  → local validation, type-check, preview, and manifest entry
  → register tasks and run the root workflow in process
  → read the final state and result snapshot printed by orchard run

Kind	Start here
Tutorials	First run
How-to	Environment setup, Run a harness, Generate a workflow with CLI, Inspect a local run, Design token-efficient schemas, Write tests, Handle errors, Troubleshooting
Reference	Packages, Cursor harness, Codex harness, Claude harness, Devin harness
Explanation	Architecture