AI Memory Systems a Feature Comparison

Live comparison table: carsteneu.github.io/ai-memory-comparison

86 systems, 79 features, source-backed. Every ✅ links to the exact code or docs that prove it.

What this is

A feature-level comparison of memory systems for AI coding agents. 86 systems compared across 79 features on seven axes: Data Model, Search & Retrieval, Knowledge Lifecycle, Extraction Pipeline, Platform Support, Architecture, and Benchmarks. Every claim backed by a public source citation — no marketing, no inference.

Star History

Quick start

• Browse the table: carsteneu.github.io/ai-memory-comparison — sortable, filterable, underdogs first
• Read the evidence: evidence/ directory, one file per system, citations to specific source lines
• Contribute: See CONTRIBUTING.md

Methodology

Every ✅ requires at least one public source: README, docs, or source code. If a feature isn't documented, it's marked ❌. Code beats docs — if docs claim a feature but the implementation doesn't exist, it's ❌. Systems are re-auditable at any time via their evidence files.

Curation criteria

1. Specifically designed for AI agent memory — not general vector DBs, not RAG frameworks
2. Intended for coding agents — or adaptable to them (Claude Code, Codex, OpenCode, etc.)
3. Open source — public repo with a recognized license
4. Not abandoned — systems inactive for extended periods may be removed

Contributing

1. Corrections: Open a PR with a link to the public source proving the correction
2. New systems: Submit a PR with a complete evidence file — see CONTRIBUTING.md for details
3. New features: PR with sources for ALL systems in that row

Project maintainers: corrections to your project's row get priority review.

License

MIT — attribution required. Copy it, fork it, use it anywhere.

Disclosure

This comparison is maintained by the author of YesMem. YesMem is listed alongside every other system and follows the same evidence rules.