recall-loom
Project memory for long-running AI work across agents, models, and sessions. Keep context, decisions, progress, and next steps in local project files.
What is recall-loom?
Project memory for long-running AI work across agents, models, and sessions. Keep context, decisions, progress, and next steps in local project files.
Overview of recall-loom
What it is
RecallLoom is a file-based project memory layer for AI collaboration. It saves context, progress, key decisions, and next steps into Markdown/JSON files stored alongside the project. No database, RAG, or background service is required. It helps maintain continuity across different sessions, models, and AI tools, so the next AI interaction can pick up the current state without re-explaining the project from scratch.
Key features
- Low restart tax: when switching sessions, models, or returning after days, the project state is restored first - Faster handoff: current summary, recent progress, and next steps are read first to quickly get into context - More efficient context usage: small and accurate project memory is read first; deep dives happen only when needed - Cross-tool handoff: project facts stay with the workspace when switching models, sessions, or AI tools - More stable writes: progress recording, current summary, and validation actions have clear paths, reducing the risk of writing outdated facts back - File-based persistence: memory is stored in readable, reviewable, and migratable Markdown/JSON files
Use cases
- Long-running software projects: let AI tools pick up current implementation state, key decisions, blockers, and next steps - Product / PRD / RFC writing: preserve scope changes, decision reasons, open questions, and collaboration threads - Research writing: maintain arguments, evidence, source boundaries, and writing progress - Multi-AI-tool handoff: reduce project state loss when switching models, tools, or sessions - Private or local-first workspaces: project state stays within the workspace for easier auditing and migration
Getting started
Install RecallLoom with: bash npx skills add https://github.com/Frappucc1no/recall-loom --skill recallloom To update the installed skill later: bash npx skills update After installation, initialize the project memory by saying something like “initialize this project” or using the short trigger rl-init in a compatible AI coding tool.
License and pricing
Apache-2.0 license. Open source.
Based on the project's public README and repository metadata.