By arianjad
Persistent knowledge vault for Claude Code — brain/ synthesis layer, 5 lifecycle hooks, vault scaffold, Nexus search integration, Remember continuity. Domain-agnostic.
Run a vault health check — frontmatter compliance, broken links, orphan notes, stale inbox, brain/ coverage.
Capture the current context to the vault — explicit save with full provenance (curated tier).
Initialize the claude-memory vault — folders, brain/, templates, Home.md, vault-manifest.json. Safe to re-run.
End-of-session reflection — review the session for vault-worthy knowledge and triage inbox notes.
Morning briefing — orient on goals, active projects, recent decisions, open issues. Picks up where the last session ended.
Use this agent to load all vault context about a specific topic — person, project, concept, or system. Gathers notes, backlinks, mentions, and produces a synthesized briefing. Invoke when deep context is needed beyond what SessionStart provides. <example> Context: User asks about a project they haven't touched in weeks user: "What do we know about the RaX beam simulation?" assistant: "I'll dispatch the context-loader to gather all vault knowledge about RaX beam simulation." <commentary>Deep context needed beyond SessionStart — dispatch context-loader for full briefing.</commentary> </example>
Use this agent to review a session for vault-worthy knowledge. Dispatched at session end by the reflect skill. Analyzes the session for solutions discovered, decisions made, constants identified, and gotchas encountered. Returns structured capture proposals. <example> Context: End of a debugging session user: "wrap up" assistant: "I'll dispatch the reflector to review this session for knowledge worth capturing." <commentary>Session ending triggers reflect skill, which dispatches reflector agent.</commentary> </example>
Use this agent to run a full vault maintenance audit. Dispatched by the /claude-memory:audit skill. Checks for orphan notes, broken wikilinks, frontmatter completeness, stale project notes, brain/ health, and inbox size. Writes a dated report and returns a top-5 summary. <example> Context: /claude-memory:audit skill dispatches this agent user: "audit vault" assistant: [dispatches vault-librarian] <commentary>Vault-librarian runs all checks and summarizes findings without auto-fixing anything.</commentary> </example>
Use this skill when the user says "triage inbox", "clean up inbox", "sort the inbox", "/claude-memory:triage-inbox", or when a reflect or wrap-up run finds more than 5 notes in 00-inbox/ and auto-invokes it. Presents one action (promote / merge / link / discard) per inbox note for batch approval, then executes approved actions. <example> Context: Inbox has accumulated 8 captures over the week. user: "triage inbox" assistant: "Loading all inbox notes and proposing one action each." <commentary>Trigger phrase matches — invoke this skill, not reflect.</commentary> </example>
This skill should be used when the user says "wrap up", "end session", "session done", "close out", "done for today", "let's stop here", or "/claude-memory:wrap-up". Performs a session-end review: verifies note quality, updates brain/ accumulators, checks for orphaned notes, promotes stale thinking/ notes, and commits all vault changes.
This skill should be used when the user wants to run a vault health check, audit vault contents, check vault health, or perform vault maintenance. Trigger phrases: "audit vault", "check vault health", "vault maintenance", "/claude-memory:audit". <example> Context: User notices vault has grown and wants to check its health user: "audit vault" assistant: "I'll dispatch the vault-librarian to run a full health check." <commentary>Trigger phrase matches — dispatch vault-librarian agent and present findings.</commentary> </example>
This skill should be used when the user asks for a "briefing", says "orient me", "catch me up", "what was I working on", "give me a status update", or when full project context is needed at the start of a session. Reads brain/ compiled knowledge first, then Remember timeline, then vault notes for a structured orientation summary.
This skill should be used when the user says "dump this", "capture this", "vault this", "remember this", "save this", "log this", "note this", "/claude-memory:dump", or provides a freeform block of information (decisions, solutions, issues, ideas, references, project updates) and wants it routed into the vault with proper structure. Accepts raw unstructured input and handles all classification, placement, and frontmatter automatically.
Modifies files
Hook triggers on file write and edit operations
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A persistent knowledge vault plugin for Claude Code. Every Claude Code session starts with a blank context — no memory of what you debugged last week, what decisions you made, or what gotchas you already discovered. This plugin gives Claude a growing knowledge base it reads at session start and writes to continuously, so knowledge compounds across sessions instead of resetting.
Claude starts every session from scratch. The Remember plugin captures a raw event stream, but raw transcripts are not the same as compiled knowledge — they are noise with signal buried in them. You end up re-explaining context, re-discovering the same solutions, and repeating mistakes you already fixed.
This plugin maintains a vault of compiled, structured knowledge that Claude reads at session start. New findings get distilled into the vault during and after each session. The signal accumulates; the noise stays out.
Remember (raw stream) --> Reflect (session end) --> brain/ (compiled) --> SessionStart (orientation)
session-orient.sh, which loads brain/goals.md and brain/skills.md into context and prints a one-line orientation summary.The vault itself is a folder of plain Markdown files in an Obsidian vault. No cloud services, no lock-in.
claude plugin install claude-memory
Or load from a local clone:
claude --plugin-dir /path/to/claude-memory
Then, in your first session with Claude Code:
/claude-memory:init
/claude-memory:standup
/claude-memory:init scaffolds the vault folder structure and copies templates. /claude-memory:standup reads brain/ and gives you a briefing on current goals and recent context.
Prerequisites:
Optional (recommended):
claude mcp add nexus -- node <vault>/.obsidian/plugins/nexus/connector.jsclaude plugin install rememberThe plugin degrades gracefully when Nexus or Remember are unavailable.
| Phase | Action |
|---|---|
| Session start | Run /claude-memory:standup — get a briefing on goals and recent context |
| During work | Ambient capture happens automatically via hooks — no action needed |
| Explicit capture | Say "vault this" or "remember this" to capture something specific |
| Freeform capture | Run /claude-memory:dump to brain-dump notes, links, or observations |
| Session end | Run /claude-memory:wrap-up — prompts wrap-up checklist and updates brain/ |
| Between sessions | brain/ persists; next standup reads it cold |
| Hook | Trigger | What it does |
|---|---|---|
SessionStart | Session opens | Runs session-orient.sh: reads brain/goals.md and brain/skills.md, prints orientation summary |
UserPromptSubmit | Each user message | Runs classify-message.py: detects 9 signal categories (DECISION, SOLUTION, GOTCHA, FUTURE, etc.) and injects routing hints for vault note creation |
PostToolUse | After Write or Edit | Runs validate-write.py: checks that new vault notes have required frontmatter fields for their tier |
PreCompact | Before context compaction | Runs pre-compact.sh: backs up the current transcript excerpt to thinking/ so Reflect can read it after compaction |
Stop | Session ends | Prints the session end checklist: run wrap-up, run reflect, check for orphan notes |
The brain/ folder holds compiled knowledge accumulators. Claude reads and updates these, not the user. Do not manually edit unless you want to seed initial content.
| File | What it stores |
|---|---|
brain/index.md | Recent context: last session summary, active threads, open questions |
brain/goals.md | Current focus areas and priorities. Read at every session start. |
brain/decisions.md | Log of non-obvious choices with rationale and date. Linked to full decision notes. |
brain/patterns.md | Recurring patterns Claude has observed across sessions |
brain/gotchas.md | Known failure modes, surprises, and things that bite. Consulted before major work. |
brain/skills.md | Available skills, agents, and tools Claude knows about for this vault |
During /claude-memory:reflect and /claude-memory:wrap-up, Claude appends new entries to the relevant accumulators and updates brain/index.md with a session summary.
npx claudepluginhub arianjad/claude-memory --plugin claude-memoryMulti-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Curate auto-memory, promote learnings to CLAUDE.md and rules, extract proven patterns into reusable skills.
Memory compression system for Claude Code - persist context across sessions
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.