Decision Log for Agents

Turn architecture decisions and past findings into agent-retrievable context

Problem

When coding with AI, the agent doesn't know why things were designed this way or which approaches already failed, so it keeps proposing rejected ideas. These decisions are scattered across PRs, issues and personal notes, unusable by the agent.

Solution

A lightweight decision-record library: structured templates for context, options, conclusion and affected scope, plus an MCP/CLI interface so coding agents can retrieve relevant decisions before editing. Commits can be linked to affected decision entries.

Users

Small engineering teams and solo developers coding with Claude Code/Cursor-style agents

MVP

Editor and search for structured decision entries; file-based storage inside the Git repo; MCP/CLI query interface; code path to decision mapping; PR template reminder to log new decisions

Revenue

Per-seat team SaaS subscription (hosted + retrieval service) or self-hosted license; free tier limited to a single local library

Why now

With agentic coding, missing context is the new bottleneck, and existing docs tools weren't designed to be read by agents; simple storage and retrieval fit one developer starting with a CLI plus MCP