MIF
active
Tech Stack:
JSON-LD
Obsidian
W3C PROV
Markdown
MIF defines a standard way for AI coding assistants to store and exchange persistent memory. Today every tool invents its own format: Claude Code writes memories one way, Cursor another, Copilot yet another. Moving between tools means losing context. MIF fixes this with a dual-format design.
Key Features
- Dual Format — Markdown frontmatter for human reading (works natively in Obsidian) and JSON-LD for machine interoperability
- Three Memory Types — Semantic (facts and knowledge), episodic (events and experiences), procedural (workflows and how-to)
- Bi-Temporal Tracking — Records both when something happened and when it was captured, enabling accurate temporal reasoning
- W3C Provenance — PROV-O compatible tracking of where memories originate and how they transform
- Conformance Levels — Level 1 (basic storage), Level 2 (full interchange), Level 3 (federation across systems)