claude-memory-kit

Solid

Persistent memory for Claude Code agents with an agent-audit-ritual architecture. User only talks; the agent captures, audits, proposes promotions, and writes. Memory lives in layers — a hot cache (MEMORY.md) held under three size caps, per-session handoffs (context/handoffs/), topical knowledge articles (knowledge/concepts/), and canonical rules (.claude/rules/) — plus multi-project isolation via projects/<name>/ and an experiments/ sandbox. /close-session runs the end-of-session audit ritual. Zero external dependencies.

AI & Automation 27 stars 7 forks Updated 1 weeks ago MIT

Install

View on GitHub

Quality Score: 87/100

Stars 20%
48
Recency 20%
90
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
50
License 10%
100
Description 5%
100

Skill Content

# Claude Memory Kit v5 Persistent memory for Claude Code agents. Layered memory, agent-driven promotion, zero manual file editing. The v5 **lean core**: two operators, per-session handoffs, memory that fails loudly instead of rotting quietly. ## Two core invariants 1. **User only talks. Agent captures, proposes, writes.** Every architectural decision passes this test. 2. **Every memory entry carries a `[YYYY-MM-DD]` date tag.** This is what lets `/close-session` detect cross-session repetition and propose promotions. ## What's new in v5 - **Session close is `/close-session`, not `/close-day`.** The ritual: capture dated patterns → audit for 3+-date repetition → promote on your "yes" → replace the MEMORY.md header (current state, never a chronicle) → write a per-session handoff. - **Per-session handoffs replace the rolling next-session-prompt.** One immutable note per closed session in `context/handoffs/<topic>-YYYY-MM-DD.md`; the SessionStart hook injects the newest one. No rolling file to freeze while looking authoritative. - **Three independent MEMORY.md caps, hook-enforced: 180 lines / 32 KB / 3000 chars per line.** Line count alone lies — content densifies into ever-longer lines while `wc -l` stays flat. Any tripped cap opens the next session with an audit prompt. - **Stale-reference detector.** Every session start, file paths mentioned in memory are checked against disk; anything moved or missing is flagged. - **The daily journal (`daily/` + `/close-day`) moved to o...

Details

Author
awrshift
Repository
awrshift/claude-memory-kit
Created
4 months ago
Last Updated
1 weeks ago
Language
Python
License
MIT

Integrates with

Anthropic · AI

Similar Skills

Semantically similar based on skill content — not just same category

AI & Automation Listed

ccc-memory

Two-tier persistent memory for Claude Code sessions. Layer 1 is CLAUDE.md (always-loaded context). Layer 2 is a memory/ directory with dated notes, decisions, and…

6 Updated today
KevinZai
AI & Automation Solid

add-persistent-project-memory-and-hook-driven-synthesis-with-cla

Keep Claude Code sessions grounded in prior decisions, project context, and daily handoff notes instead of starting from zero every time.

20 Updated today
agentskillexchange
AI & Automation Listed

bettermemory

Verification-grade memory between sessions. Use bettermemory's MCP tools (memory_search, memory_show, memory_write, memory_verify, memory_record_use, etc.) instead of writing to files when the user asks you to "remember" something or references shared context from a past session. Default is to NOT call memory_search; only retrieve when the user references context you don't have ("my project", "the script we wrote") or a request is ambiguous in a way stored preferences could resolve. Every hit carries a staleness_verdict (calendar + path-drift + commit-drift); when it isn't "fresh", spot-check a claim before relying and call memory_verify to attest. Every use should record a claim_excerpt so retrievals stay auditable.

0 Updated today
0Mattias