Memory Interchange Format (MIF)

An open standard for portable AI memory representation.

Overview

MIF is a proposed standard that aims to enable interoperability between AI memory providers by defining a common data model with dual representations:

Markdown (.memory.md) - Human-readable, Obsidian-compatible
JSON-LD (.memory.json) - Machine-processable, semantically linked

Why MIF?

The AI memory ecosystem is fragmented. Mem0, Zep, Letta, LangMem, Subcog, and others all use proprietary schemas. MIF solves:

Vendor lock-in - Move memories between providers
Data ownership - Local-first, plain text files
Interoperability - Common vocabulary for relationships and entities
Future-proofing - Standard formats survive provider discontinuation

Quick Start

Minimal Memory (Markdown)

---
id: 550e8400-e29b-41d4-a716-446655440000
type: semantic
created: 2026-01-15T10:30:00Z
namespace: _semantic/preferences
---

User prefers dark mode for all applications.

Minimal Memory (JSON-LD)

{
  "@context": "https://mif-spec.dev/schema/context.jsonld",
  "@type": "Memory",
  "@id": "urn:mif:550e8400-e29b-41d4-a716-446655440000",
  "memoryType": "semantic",
  "namespace": "_semantic/preferences",
  "content": "User prefers dark mode for all applications.",
  "created": "2026-01-15T10:30:00Z"
}

Key Features

Feature	Description
Dual Format	Markdown AND JSON-LD representations
Obsidian Native	Valid Obsidian notes with wiki-links
Typed Relationships	9 core relationship types (RelatesTo, DerivedFrom, Supersedes, etc.)
Entity Types	Person, Organization, Technology, Concept, File
Bi-Temporal	Track when recorded AND when facts are valid
Flexible Namespaces	`_public`, `_shared`, `{org}/{scope}+` with reserved prefixes
Model-Agnostic Embeddings	Store model + source text for re-embedding
Citations	Structured references with type/role taxonomy (Level 3)
W3C PROV	Standard provenance tracking
JSON Schema	Automated validation for MIF documents
Ontology System	Three-type hierarchy with domain extensions

Ontology System

MIF includes an ontology system for organizing memories into three base types:

semantic/              # Facts, concepts, relationships
├── decisions/         # Architectural choices, rationale
├── knowledge/         # APIs, context, learnings, security
└── entities/          # Entity definitions

episodic/              # Events, experiences, timelines
├── incidents/         # Production issues, postmortems
├── sessions/          # Debug sessions, work sessions
└── blockers/          # Impediments, issues

procedural/            # Step-by-step processes
├── runbooks/          # Operational procedures
├── patterns/          # Code conventions, testing
└── migrations/        # Migration steps, upgrades

Features

Hierarchical namespaces - Cognitive memory types with sub-namespaces
Entity types - Define custom entities with traits and schemas
Discovery patterns - Auto-detect capture opportunities from content
JSON-LD export - Semantic web compatibility via yaml2jsonld.py
Extensible - Domain-specific ontologies extend the base

See ontologies/ for base ontology and examples.

Specification

See SPECIFICATION.md for the complete technical specification.

Validation

JSON Schema files are available for automated validation:

# Validate a MIF document
npx ajv validate -s schema/mif.schema.json -d your-memory.json

# Validate citations only
npx ajv validate -s schema/citation.schema.json -d citation.json

# Validate ontology definitions
npx ajv validate -s schema/ontology/ontology.schema.json -d ontology.yaml

Ontology Conversion

Convert YAML ontologies to JSON-LD for semantic web compatibility:

python scripts/yaml2jsonld.py ontologies/mif-base.ontology.yaml
python scripts/yaml2jsonld.py --all  # Convert all ontologies

Examples

See the examples/ directory for:

Minimal memories (Level 1)
Standard memories with relationships (Level 2)
Full-featured memories with temporal, provenance, and citations (Level 3)

Conformance Levels

Level	Requirements
Level 1: Core	id, type, content, created
Level 2: Standard	+ namespace, entities, relationships, timestamps
Level 3: Full	+ bi-temporal, decay, provenance, embeddings, citations

Migration

MIF includes migration guides for:

Mem0
Zep
Letta (Agent File)
Subcog
Basic Memory

Contributing

This specification is open source. Contributions welcome:

Open an issue for discussion
Submit PRs for specification changes
Implement and share converters

Subcog - AI memory system implementing MIF
Mnemonic - Claude Code plugin using MIF ontologies
Issue #82 - Original proposal

Citing This Project

If you use MIF in your research or projects, please cite it:

@software{allen_mif_2026,
  author       = {Allen, Robert},
  title        = {{Memory Interchange Format (MIF)}},
  version      = {0.1.0},
  date         = {2026-01-23},
  url          = {https://mif-spec.dev},
  license      = {MIT}
}

You can also use GitHub's built-in "Cite this repository" button on the repository page to get an auto-generated citation in APA or BibTeX format.

Acknowledgments

MIF builds on the following standards:

Manu Sporny, Dave Longley, Gregg Kellogg, Markus Lanthaler, Pierre-Antoine Champin, and Niklas Lindstrom. JSON-LD 1.1: A JSON-based Serialization for Linked Data. W3C Recommendation, 2020. https://www.w3.org/TR/json-ld11/

Luc Moreau and Paolo Missier. PROV-DM: The PROV Data Model. W3C Recommendation, 2013. https://www.w3.org/TR/prov-dm/

License

MIT

MIF