Memory plugins

跨会话长期记忆工具集 — 5 个核心 skill (recall/capture/commit/doctor/admin)，支持 OpenViking / MCP / mem0 三后端，identity 默认 fail-closed。

6 skills · 3 commands

Recall at session start and per prompt, batched capture with secrets removed, StateJar's MCP tools, and /statejar commands.

6 skills · 1 MCP server · hooks

MayringCoder memory integration: hybrid retrieval, conversation summary capture, pi_task dispatch via local MCP.

4 skills · 1 command · 1 MCP server · hooks

Persistent Obsidian-based memory for Claude Code. Knowledge vault that persists across sessions — project context, architecture decisions, component notes, patterns, and TODOs.

3 skills · 2 commands · 1 agent

PowerMem intelligent memory for Claude Code: auto-recall on prompt submit and auto-save on session end via HTTP hooks.

4 skills · hooks

Hydra Claude: memory provider — auto-captures repo-specific patterns at session end and re-injects them on session start and post-compact.

3 skills · hooks

Local-first MCP server that gives Claude Code a living, browser-viewable project wiki and a project-local memory store.

1 skill · 1 MCP server · hooks

Verification-grade memory for Claude Code. Every retrieved fact carries a staleness verdict (calendar age + filesystem path drift + git commit drift) so the model knows when a stored memory has rotted before relying on it; retrievals are logged with three attribution tiers (model-explicit with claim-level excerpts, Stop-hook substring match, or auto-fallback) so a reply months later can be traced back to the load-bearing stored claim wherever it was recorded. Memories are plain markdown + YAML in ~/.claude-memory/; retrieval is opt-in per turn, with a score-gated recall hook injecting a single stored pointer on the ~2% of prompts the silent-miss probe flags; writes about you always stage for confirmation. A sibling episode tier (5 episode_* tools) gives /loop iterations and subagent handoffs a home for journal-shaped run-state the durability gate rejects. Bundles 27 MCP tools — 18 registered by default, the nine curation/power-user tools behind full_tool_surface (22 memory_* + 5 episode_*) — and a memory-discipline skill that lands the policy in the system prompt.

1 skill · 1 MCP server · hooks

Shared memory graph: MCP tools (recall/save/find/link) + 3D context universe web UI. All agents, one brain.

1 skill · 1 command · 1 MCP server

Long-term memory for Claude Code: recall before work, remember after, with forgetting and procedural version control.

1 skill · 1 MCP server

Sovereign AI memory with Crystallization Theory governance. Human-gated crystallization, δ-decay, witness chains, freeze/revert. One skill + one MCP server = governed memory for any agent.

1 skill · 1 MCP server

Persistent session memory for Claude Code backed by an Obsidian vault exposed through an MCP filesystem server. Provides /resume and /save with first-run bootstrap that registers the vault MCP on each new machine.

2 skills

Persistent cross-session memory for AI agents. Two storage modes: semantic vector search (mem_*) for knowledge found by meaning, and key-value store (kv_*) for instant lookup. Auto-expiry via TTL + volatile-lru eviction.

1 skill · 1 MCP server

Search your markdown notes and recall what past AI sessions decided, all on your own machine.

1 MCP server · hooks

SQLite-backed long-term memory with tiers (working/episodic/semantic), W3C Verifiable Credentials, drift detection, and wing/room access control. The memory + credential layer of the Claw Stack.

1 skill · hooks

Use when designing or reviewing how an AI agent stores, organizes, and reuses what it learns across tasks and sessions.

2 skills

Use when testing whether agent memory helps or harms task outcomes, or when diagnosing wrong, stale, or missing recall.

2 skills

Cross-session memory for Claude Code. A curated store of markdown files under ~/.eve, recalled automatically on every prompt. POSIX sh and awk; no dependencies, no network, no telemetry.

hooks

Trusted engineering context for AI coding agents. Keeps Claude Code and Codex sessions current as the code changes, with checks before edits and scoped corrections.

1 MCP server

Persistent persona memory. Mesh-structured state (identity + voice + decisions + corrections + verbatim log) that survives context compaction. Project-isolated.

1 skill

Use when an agent needs to save where it is in a long task so it can resume later, or when handing a task to another agent without losing context.

1 skill

Bridge between ai-iq smart memory (semantic + graph + FSRS) and Anthropic Managed Agents memory_stores.