oh-my-dag
Active·★ 44·MIT·Updated 2026-10-11
Oh-my-dag (omd) serves as an orchestration layer beneath your AI coding agent, transforming conversational tasks into structured, checkpointed DAGs. It enforces rigorous verification through exit codes, cross-model critiques, and artifact gates, ensuring the reliability and correctness of agent-generated work.
#Orchestration#AI Agent#Workflow Automation#DAG (Directed Acyclic Graph)#Code Verification#LLM Ops#Pipeline#Task Management
01
Features
01Exit Code as Finish Line: Tasks conclude when a declared command exits successfully, preventing models from self-certifying completion.
02Rigorous Criterion Verification: Acceptance tests run against pre-work and deliberately wrong artifacts to ensure they genuinely test the solution.
03Multi-Model Orchestration: Allows different LLMs for each node in a DAG, enabling tailored model selection and cross-family second opinions for critical tasks.
04Resumable & Checkpointed Workflows: Every completed node is atomically saved to disk, allowing interrupted work to resume from the last checkpoint, saving time and billing.
05LLM-Free Retrieval: `omd_web` performs web searches and fetches without an LLM in the loop, ensuring factual accuracy based on retrieved content rather than model memory.
02
Compatibility
Bun
Runtime
Verified via docs
03
Quick start
1
$ git clone https://github.com/AbyssCN/oh-my-dag.git
2
$ cd oh-my-dag
3
$ bun install
4
$ bun link
5
$ omd init
04
Use cases
↳Deep Research & Analysis: Conduct multi-stage research pipelines involving seeded crawls, fan-out lenses, judge panels, and gap rounds using different models for each stage.
↳Automated Design Argumentation & Specification: Argue design choices and then generate executable specifications (e.g., `/omd-grill` followed by `/omd-contract`).
↳Cross-Model Code Review: Perform multi-dimensional code reviews, falsifying findings across different models to ensure comprehensive and unbiased feedback (`/omd-review`).
↳Automated Debugging Workflows: Systematically reproduce issues, lock scope, test parallel hypotheses, and verify fixes (`/omd-debug`).
↳Controlled Delivery of Decision Maps: Manage decision maps in Git, ruled by user-defined triggers, preventing unattended merging (`/omd-deliver`).
05
Alternatives
tesseron★ 21
Tesseron protocol, documentation, MCP gateway and Claude Code plugin. Created and maintained by Eigenwise.
api★ 15
Agent skills + a chat MCP server to operate the ibl.ai platform via its REST API. Install: npx skills add iblai/api
architect-to-product★ 11
Vibe code fast. Ship like an engineer. AI engineering framework packaged as an MCP server that enforces TDD, security scanning, pentesting, secret management, QA audits, and deployment to a securely configured VPS — automated with up to 100× fewer exploration tokens.
nac★ 284
Give AI agents ambitious work without losing the plot. nac is an open-source harness for long-running tasks, using a central orchestrator, threads, and structured episodes to stay aligned with your intent.
Comments
Log in to leave a comment
No comments yet. Be the first!