Claude Code Memory Plugin

The Pensyve plugin gives Claude Code persistent memory across sessions. It remembers your preferences, project decisions, debugging approaches, and codebase knowledge — so you don't have to repeat yourself.

Install

In Claude Code:

/plugin marketplace add /path/to/pensyve/integrations/claude-code
/plugin install pensyve@pensyve

Then restart Claude Code.

Connect to Pensyve

Cloud (recommended)

Set your API key (get one at pensyve.com/settings/api-keys):

export PENSYVE_API_KEY="psy_your_key"

Add to your shell profile (~/.bashrc, ~/.zshrc) to persist across sessions.

The plugin ships pre-configured for Pensyve Cloud — it reads PENSYVE_API_KEY from your environment and passes it as a Bearer token in the Authorization header on every MCP request. Once the env var is set, you're ready to go.

To override the MCP config explicitly (e.g., for a specific project), add to .claude/settings.json:

{
  "mcpServers": {
    "pensyve": {
      "type": "http",
      "url": "https://mcp.pensyve.com/mcp",
      "headers": {
        "Authorization": "Bearer ${PENSYVE_API_KEY}"
      }
    }
  }
}

Use headers with Authorization: Bearer, not env. For HTTP-transport MCP servers, the headers block sends your key on every request. The env block is for local stdio servers that read environment variables at startup.

Local (self-hosted)

Build the MCP binary:

git clone https://github.com/major7apps/pensyve
cd pensyve
cargo build --release -p pensyve-mcp

Override the MCP config in .claude/settings.json:

{
  "mcpServers": {
    "pensyve": {
      "command": "pensyve-mcp",
      "args": ["--stdio"]
    }
  }
}

No API key needed — all data stays on your machine in SQLite.

What You Get

Slash Commands

Command	Description
`/remember <fact>`	Store a fact, decision, or pattern
`/recall <query>`	Search memories by semantic similarity
`/forget <entity>`	Delete all memories for an entity
`/inspect [entity]`	View all memories grouped by type
`/consolidate`	Trigger memory consolidation cycle
`/memory-status`	Show namespace statistics

Skills

Skill	When to Use
`session-memory`	End of a work session — captures decisions and outcomes
`memory-informed-refactor`	Before refactoring — loads relevant prior context
`context-loader`	Session start or context switch — loads historical context
`memory-review`	Periodic — finds stale facts, contradictions, cleanup opportunities

Agents

Agent	Mode	Purpose
`memory-curator`	Background	Monitors sessions, suggests memorable events
`context-researcher`	On-demand	Deep memory search, returns structured briefings

Hooks

Event	Behavior
`SessionStart`	Loads relevant memories at session start
`Stop`	Extracts decisions/outcomes after task completion
`PreCompact`	Persists in-flight context before compaction
`UserPromptSubmit`	Enriches prompts with memory context (opt-in)

How It Feels

First session:

You: "We're using Drizzle ORM with Postgres, deployed on Vercel" Claude remembers this.

Next session (days later):

You: "Add a new table for audit logs" Claude recalls your ORM choice and generates Drizzle schema — no re-explanation needed.

Across sessions:

Your debugging preferences persist
Architecture decisions are remembered
Code style choices carry forward
Project-specific patterns are recalled automatically

The plugin works with any MCP-compatible client, not just Claude Code. See MCP Setup for other clients, or Remote MCP for cloud setup.

On this page