
DevChronicle MCP is a local Model Context Protocol (MCP) server built for IBM Bob that introduces persistent project memory into AI-assisted development. It aggregates structured context from development logs (devlogs), Architecture Decision Records (ADRs), repository files, documentation, and git history to give AI a deeper understanding of how a project evolves over time. With this context, DevChronicle enables: Project Awareness — It provides a clear view of project goals, completed work, key decisions, and pending tasks, allowing IBM Bob to operate with full situational context. Change Risk Analysis (Core Feature) — Before any modification, DevChronicle evaluates potential risks by detecting conflicts with prior decisions, ADRs, and critical system components. This prevents unsafe or inconsistent AI-generated changes. Persistent Engineering Memory — Development progress is continuously recorded as structured Markdown, capturing tasks, decisions, blockers, and next steps—creating a reliable, machine-readable project history. Architecture Decision Tracking — DevChronicle generates and maintains ADRs, ensuring that important technical decisions are documented, traceable, and respected in future development. Feature Recommendations — Based on the current project state, it suggests next steps, improvements, and enhancements across reliability, testing, documentation, and product evolution. How It Works DevChronicle is designed to be deterministic and lightweight. The MCP server performs structured analysis of local files and repository data without relying on external LLM APIs. Instead, IBM Bob’s inbuilt model interprets the structured outputs, combining reasoning with reliable context. Communication is handled via STDIO using MCP, ensuring a clean separation between data extraction (DevChronicle) and intelligent reasoning (IBM Bob). This approach guarantees reliability, low cost, and explainability, making it suitable for real-world engineering workflows.
17 May 2026