MultiAgentCoding

Visual workflows and an execution control plane for AI coding agents

Active Development

75%

Description

A Python/FastAPI control plane for configuring and running multi-agent coding workflows. It combines model-independent agents, reusable roles, skills, prompt profiles, evaluation rubrics and a visual workflow designer with an Obsidian knowledge vault. The documented runtime uses OpenCode with execution planning, dry runs, cancellation and bounded retries. Direct provider adapters and persistent run storage remain planned work.

Architecture

Agent Dashboard / Workflow Designer / Terminal → Workflow Planner → OpenCode Adapter → Execution Runtime → Events and Results → Obsidian Knowledge Vault.

Tech Stack

PythonFastAPIJavaScriptOpenCodeAI AgentsWorkflow GraphsObsidianBYOK

Implementation Roadmap

Implement agent identities, reusable roles, skills and prompt profiles.

Build the dashboard, workflow designer and terminal interfaces.

Connect controlled execution, evaluation and persistent project knowledge.

Add direct provider adapters and persistent run storage.

GitHub