# claude-memory

claude-memory keeps past Claude sessions in a local SQLite file, then hands the most recent one back as context when a new session begins.

- plugin · `type=plugin` · Claude Code · ★ 276 repo stars
- Author: Samarth Gupta
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/gupsammy/Claudest/tree/HEAD/plugins/claude-memory
- Updated: 2026-10-29
- Security check: automated, 2026-11-16 (text and 33 script files) — warning: a command downloads a script from ollama.com and runs it right away (file plugins/claude-memory/hooks/semantic_setup.py). Do not run such commands without an explicit yes from the human.
- HTML: https://affpapa.org/en/nevibecode/plugins/claude-memory

## What it does

A stop hook reads the session transcript in JSONL, splits it into messages, marks branches left by rewinds, and writes it all to `~/.claude-memory/conversations.db`. On the next start a second hook pulls the latest substantial session for the same project and injects it. That is what makes plan-now, build-later survive a context reset.

Two skills ship with it. `recall-conversations` does keyword lookup through FTS5 with BM25 ranking, or browses sessions in time order, and the agent writes the queries itself from your wording. `extract-learnings` rereads old chats, picks out insights worth keeping, and suggests which memory layer each one belongs to (`CLAUDE.md`, `MEMORY.md`, topic files). Nothing is written until you approve.

Python 3.7+ is the only prerequisite, standard library included. Semantic search is opt-in: EmbeddingGemma 2 runs inside a local Ollama, the model download is about 360 MB, and the switch is `python3 hooks/semantic_setup.py --enable`. With Ollama down, search falls back to keywords and warns you. `semantic_min_score` sits at 0.70 by default and drops meaning-based hits below that similarity.

## Install

### Claude Code (native platform)

**Plugin marketplace**
In the agent chat (not a shell):
```text
/plugin marketplace add gupsammy/Claudest
```
In the agent chat (not a shell):
```text
/plugin install claude-memory@Claudest
```

## When to use it
- resume a project from the point the last session stopped
- find the chat where the schema migration came up
- see which files were touched last time
- move lessons from old sessions into `CLAUDE.md`
- search past conversations by meaning rather than exact wording

## Example prompts
- What did we settle on for the database migration last session?
- Pick up where we left off yesterday.
- Dig through our past chats and tell me why we dropped that approach.

## Similar
- [Titen Memory](https://affpapa.org/en/nevibecode/mcp/titen-memory.md) — MCP server · ★ 19 repo stars
- [Local Memory](https://affpapa.org/en/nevibecode/mcp/local-memory.md) — MCP server · ★ 15 repo stars
- [recall](https://affpapa.org/en/nevibecode/plugins/recall.md) — plugin · ★ 23 repo stars
- [claudex](https://affpapa.org/en/nevibecode/mcp/claudex.md) — MCP server · ★ 97 repo stars
- [wiki-query](https://affpapa.org/en/nevibecode/skills/rohit-ghumare-wiki-query.md) — skill · ★ 2966 repo stars
- [replay](https://affpapa.org/en/nevibecode/commands/replay.md) — command · ★ 2966 repo stars

From the same set: [gupsammy/claudest](https://affpapa.org/en/nevibecode/by/gupsammy/claudest.md) — 47 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
