Agent Memory Doctor

Diagnose and fix memory recall issues for long-term coding agent users

Problem

Developers increasingly rely on agents like Hermes as daily assistants, but built-in memory plugins have low recall accuracy and weak relevance, and alternatives like hindsight are hit-or-miss. Meanwhile Skills, MCP, and AGENTS.md configs pile up across agents, making it unclear what each agent actually loads, with duplicate skills, broken symlinks, and corrupted configs common.

Solution

A local tool that scans each agent's memory store and configs to produce a Memory Health Report: which memories are never recalled, which recalls are irrelevant, and which configs are duplicated or broken. It offers one-click backup, rollback, and re-indexing, and exposes diagnostics so the coding agent itself can apply fix suggestions.

Users

Indie developers and power users running Claude Code, Codex, Cursor, and Hermes side by side

MVP

Support scanning memory directories for 2–3 mainstream agents; output recall hit-rate and duplicate/stale entry lists; one-click backup and rollback; generate paste-ready fix instructions for the agent

Revenue

Free scan with paid repair/automation subscription, or a one-time purchase via agent plugin marketplaces; later add team-shared memory governance

Why now

Coding agents are moving from experimentation to daily dependency, and memory/config fragmentation is just becoming widely felt; big vendors are not yet addressing cross-agent local governance