StackMap
Subscribe
Explore / deer-workflow
deerwork-ai

deer-workflow

Graph-engineering runtime where TypeScript owns control flow and replaceable coding-agent CLIs — Codex, Claude Code, Pi — do the semantic work inside each node. Pilot for DeerFlow 3.0.

487 52 TypeScript MITupdated 15 days ago
orchestrationcoding
View on GitHubDispute this mapping →
Curator's take

Pick it when the orchestration itself has to be reviewable: phases, inputs, retries and failure handling live in TypeScript you diff in a PR, not in a prompt that hopes the agent remembers the plan. The generator is the hook — `deer-workflow create "<description>"` has Codex apply the bundled workflow-creator skill and prints a runnable module you then edit as normal code, and `--print` streams one JSON event per line so CI can consume a run. Caveats worth knowing before adopting: Bun-only, requires an installed and signed-in agent CLI, ships two examples, and is explicitly a pilot for ByteDance's DeerFlow 3.0 — expect the API to move.

Mapped by ShipWithAI editors · links verified
README.md

English | 简体中文

deer-workflow

License: MIT npm Bun TypeScript Codex CLI DeerFlow Stars GitHub Stars

An open-source Dynamic Workflow runtime for building observable, reusable Agent graphs.

deer-workflow is a pilot project for DeerFlow 3.0, also known as DeerWork.

Index

Why Deer Workflow

Deer Workflow is a code-first implementation of Graph Engineering: TypeScript defines the valid execution paths, while Coding Agents perform the semantic work inside each node.

  • Code is the plan. Control flow, phases, inputs, and failure handling live in reviewable TypeScript rather than an opaque Agent conversation.
  • Agents are replaceable. Codex is the default runtime; Claude Code and Pi are built in; and the public Agent interface remains vendor-neutral.
  • Execution is observable. Interactive runs provide a phase-aware TUI; automation can consume a stable JSONL event stream.

How to use

Quick Start

Install Bun and sign in to Codex CLI, then install the released CLI:

bun install --global @deerwork-ai/deer-workflow

Describe the orchestration you want. Deer Workflow asks Codex to apply the bundled workflow-creator Skill and writes a runnable TypeScript module:

deer-workflow create \
  "Create a Workflow that accepts a topics string array, researches each topic in parallel, and synthesizes a report" \
  > workflow.ts

Use --agent claude or --agent pi to generate with another installed Harness. Codex remains the default.

Run the generated Workflow with its example input:

deer-workflow run ./workflow.ts \
  --input '{"topics":["Agent Skills","Dynamic Workflows"]}'

Interactive terminals show phases and Markdown logs in a live TUI. For servers, CI, and process pipelines, add --print or -p to stream one JSON event per stdout line.

Want to understand or edit the generated module? Continue with the Getting Started guide.

Examples

  • Deep Research discovers research angles, investigates them in parallel, verifies claims, and produces an interactive HTML report.
  • Blog Writer plans an article, drafts its sections through a pipeline, reviews them, and returns structured output.

These examples live in the repository. Clone or download it before running their documented commands.

Documentation

  • Getting Started — learn the execution model and build a Workflow step by step.
  • API Reference — inspect exact functions, types, events, and runtime behavior.
  • [Workflow Creator Skill](./skills/w

Continue your stack

What teams reach for next — and why each earns a place beside deer-workflow. Ranked by curator confidence.