Sup Mem — Development skill for Claude Code
Memory for Claude Code that measures whether its memories helped — outcome ledger, self-tuning retrieval, bitemporal recall, tamper-evident provenance.
How to install Sup Mem
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open kiraa06/sup-mem and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Sup Mem does
Memory for Claude Code that measures whether its memories helped — outcome ledger, self-tuning retrieval, bitemporal recall, tamper-evident provenance. SQLite, zero-infra.
Alternatives in Development
- Mempalace Search — Search your memories across the MemPalace using semantic search with wing/room filtering 58.6k ★
- AgentSkillOS — Build your agent from 200,000+ skills via skill RETRIEVAL & ORCHESTRATION 373 ★
- Recent Memories — Display the most recent memories and activity from Claude's persistent storage 332 ★
README
[](https://github.com/kiraa06/sup-mem/actions/workflows/ci.yml) [](https://pypi.org/project/sup-mem/) [](https://pypi.org/project/sup-mem/) [](https://github.com/kiraa06/sup-mem/blob/main/LICENSE) [](https://github.com/astral-sh/ruff)
   
**A self-hosted memory layer for Claude Code, Codex, and Gemini CLI that persists context across sessions — and, unlike every other memory tool, *measures whether its memories actually helped*.**
Every memory system stores and retrieves. sup-mem also closes the loop:
- Memories have a P&L. A
Stophook reads each finished response and scores every injected memory: referenced, ignored, or contradicted.sup-mem roishows what each memory costs in context tokens vs. contributes. - It tunes itself on that evidence. During development, sup-mem's ledger caught its own retrieval running at a 19% hit rate (~65k wasted tokens in one session), auto-raised its threshold losslessly (0.35 → 0.80), and a per-memory clip cut injection cost 86% — all measured, not vibes.
- Zero infrastructure by default. One command: SQLite FTS5, no Docker, no emb
Related Skills
Notary
A black-box recorder and approval gate for any Claude skill. Tamper-evident, human-approved execution receipts
Mem Rfm
Outcome-ranked agent memory in one SQLite file with Claude Code integration
Asqav Claude Code
Signed Asqav receipts for Claude Code sessions. Tamper evident evidence of what your coding agent did.
MCP Recall
mcp-recall compresses MCP tool outputs (94 KB → 3.5 KB · 96%) and stores full results in SQLite for retrieval
Skillmem
Self-improving skill memory for coding agents: learn, recall, reinforce, decay — with provenance on every memo
Mem
Use when: querying past work, managing memories, or checking project history
Related Agents
Context Retrieval
Context retrieval specialist for gathering relevant memories, code patterns, and framework documentation befor
Memory Retrieval
Searches an Obsidian vault (~/notes/) for relevant memories, context, and connections. Use before starting wor
Claude Mem
A Claude Code plugin that automatically captures everything Claude does during your coding sessions, compresse
