Bodhiorchard

Product Introduction

1. Definition: Bodhiorchard is an open-source, self-hosted Agent-Driven Development (ADD) platform and AI-powered project management alternative. It replaces traditional Agile, Scrum, Jira, and Confluence workflows with a suite of 12 specialized AI agents that automate the software development lifecycle from intake to retrospective.
2. Core Value Proposition: Bodhiorchard exists to eliminate developer busywork like story points, planning poker, stale tickets, and status meetings. By automating specs, estimates, and quality processes with AI agents, it aims to return valuable time to human developers for decision-making and creation. Its core promise is "less process, more shipped," operating under the Apache 2.0 license with a strict data plane that never leaves your machine.

Main Features

1. 12 Specialized AI Agents: Each agent handles a discrete phase of the software lifecycle.
   • How it works: The Triage Agent conducts intake interviews via Slack, deduplicates via vector search, and estimates complexity. The BUD Agent drafts a comprehensive Business Understanding Document (spec, tech plan, test plan) with codebase context. The Design Agent scopes UI/UX and component breakdowns. The Tech Plan Agent creates file-level implementation TODOs with dependency analysis. Other agents handle status updates, standup summaries, test plan generation, bug linking, and developer skill profiling.
   • Technologies: Powered by Claude Code for codebase-aware tasks, with an architecture that supports integration with other LLMs like Ollama and OpenAI. Agent actions are orchestrated via a FastAPI backend and an MCP (Model Context Protocol) server for external tool integration.
2. Predictive Cycle Time Forecasting: Replaces subjective story points with data-driven timelines.
   • How it works: It utilizes an AI-PERT model combined with Monte Carlo simulations. It runs 10,000 simulations factoring in team skill profiles, backlog size, and current workload to generate P50, P70, and P85 confidence intervals for delivery dates at each project phase.
   • Technologies: Probabilistic forecasting engine integrated into the estimation workflow, providing transparent, risk-aware scheduling.
3. Living Knowledge Base & Code Intelligence: Maintains a self-updating, searchable repository of project context.
   • How it works: A tree-sitter code-graph indexer continuously scans connected Git repositories, builds a feature-cluster map, and powers queries like blast-radius analysis. This knowledge is vector-indexed and automatically fed into every agent's prompt to prevent stale documentation.
   • Technologies: PostgreSQL with pgvector for vector search, ONNX fastembed for local embeddings, and tree-sitter for code parsing. The system auto-syncs from code to ensure information currency.
4. The BUD (Business Understanding Document) Lifecycle: Unifies all project artifacts into a single living document per feature.
   • How it works: A BUD progresses through an audited workflow: bud → design → tech_arch → development → testing → uat → prod → closed. It contains the spec, technical plan, test cases, acceptance criteria, and full history, replacing scattered tickets and docs.
   • Technologies: Implemented as a core data model in a SQLAlchemy 2.0 (async) Python backend, with a Vue 3 + Vuetify frontend (including a 3D "Living Tree" dashboard) and Colyseus for real-time multiplayer state sync.
5. Developer Skill Profiling & Smart Assignment: Matches tasks to developers based on actual expertise.
   • How it works: A Skill Agent rebuilds developer profiles nightly by analyzing Git commit history, BUD contributions, and bug fix patterns. The system then suggests the best-fit developer for a task based on per-module expertise scores and real-time capacity, moving beyond static role assignments.

Problems Solved

1. Pain Point: "Busywork overhead" in traditional Agile/Scrum environments. Teams lose productivity to manual estimation ceremonies (story points, planning poker), fragmented information across Jira, Confluence, Slack, and email, constant status meetings, and the maintenance of obsolete documentation.
2. Target Audience: Software development teams and engineering managers frustrated with Agile ritual overhead. This includes teams affected by Atlassian's sunset of self-hosted Jira (ending 2029), open-source projects seeking integrated tooling, and enterprises requiring self-hosted, auditable AI development workflows to maintain data sovereignty.
3. Use Cases:
   • End-to-End Project Management: Managing the entire lifecycle of a software feature from a Slack request to deployment and retrospective within one system.
   • Automated Specification and Test Planning: Generating initial drafts of technical specifications, UI/UX designs, and comprehensive test plans (unit, integration, e2e, security, UAT) with AI, leaving humans to review and decide.
   • Data-Driven Sprint Planning and Forecasting: Using probabilistic forecasting to set realistic deadlines and manage stakeholder expectations without subjective guessing games.
   • Cross-Repository Codebase Management: Analyzing impact and dependencies across multiple monorepos or related repositories from a centralized, intelligent dashboard.

Unique Advantages

1. Differentiation: Unlike traditional project management tools (Jira, Asana) that are manual tracking systems, Bodhiorchard is an active automation platform. Unlike AI coding assistants (Cursor, GitHub Copilot) that aid individual developers at the editor level, Bodhiorchard operates as the orchestration and management layer above the IDE, coordinating the entire team workflow. It is also fully self-hosted and AI-engine-agnostic, contrasting with cloud-only solutions like GitHub Copilot Workspace.
2. Key Innovation: The core innovation is the Agent-Driven Development (ADD) methodology, institutionalized through a suite of 12 cooperating AI agents that each own a phase of the lifecycle. This is augmented by a "Living Knowledge" architecture where the system continuously learns from the codebase and project history to inform its own agents, creating a closed-loop system that improves over time.

Frequently Asked Questions (FAQ)

1. What is Agent-Driven Development (ADD)? ADD is a software development methodology where specialized AI agents handle the busywork of specs, estimates, test plans, and triage, while humans review and steer decisions. Bodhiorchard is the open-source reference implementation of this methodology.
2. Does my source code leave my machine? No. The data plane—including your codebases, Postgres database, vector embeddings, BUDs, and audit logs—remains entirely on your self-hosted hardware. Only LLM prompts are sent externally when you choose to use a cloud-based AI provider like Anthropic. An air-gapped mode using local Ollama is on the roadmap.
3. Do I need a Claude Pro/Max subscription to use Bodhiorchard? No. You can use a direct Anthropic API key for pay-per-token inference. Alternatively, if you have a Claude Pro/Max subscription, you can run in Hybrid Mode, where the agents reuse your existing login session, combining a flat-rate subscription with the platform's automation.
4. How is Bodhiorchard different from tools like Cursor or Continue? Cursor and Continue are AI coding assistants for individual developers within their IDE. Bodhiorchard is a project management and orchestration platform for entire teams that handles intake, specs, estimation, assignment, and retrospectives. They are complementary; you can use an AI coding assistant while using Bodhiorchard to manage the overall project.
5. Can I use Bodhiorchard for my commercial, proprietary product? Yes. Bodhiorchard is released under the Apache 2.0 License, which permits commercial use, including embedding it into proprietary products. Contributions require a DCO sign-off. Separate commercial licenses with dedicated support are available from the maintainer.