@fhillipcastillo/agentic-memory

Universal memory layer for agentic AI tools (Claude Code, OpenCode, Codex, OpenClaw, etc.). Store and retrieve past conversations, solutions, conclusions, and context across sessions.

What This Is

A persistent memory system that:

• Stores conversations, solutions, conclusions, issues, context
• Retrieves via keyword or semantic search (using qmd)
• Compounds knowledge over time

Quick Start

# Install globally
npm install -g @fhillipcastillo/agentic-memory

# Initialize storage
agentic-memory init

# Setup MCP in your AI tool
agentic-memory install

# Start using in your AI tool

Installation

Global Install (Recommended)

npm install -g @fhillipcastillo/agentic-memory
agentic-memory init

npx (No Install)

npx @fhillipcastillo/agentic-memory init
npx @fhillipcastillo/agentic-memory serve

Per-Project Install

npm install @fhillipcastillo/agentic-memory
npx agentic-memory init

Setup MCP

Run the install wizard to auto-configure your AI tool:

agentic-memory install

This detects Claude Code, OpenCode, or Codex and adds the MCP config.

Manual Setup

Add to your claude.code.json, .opencode/config.json, or .codex.json:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["@fhillipcastillo/agentic-memory", "serve"]
    }
  }
}

OpenClaw Plugin

Install as OpenClaw plugin:

# Install via npm
openclaw plugins install @fhillipcastillo/agentic-memory

# Or use the install command
agentic-memory install

The plugin registers tools directly - no MCP needed:

• memory_store
• memory_get
• memory_query
• memory_context
• memory_patterns
• memory_conclusions
• memory_issues
• memory_stats
• memory_init

Manual OpenClaw Config

Add to ~/.openclaw/openclaw.json:

{
  "plugins": {
    "enabled": true,
    "allow": ["agentic-memory"]
  }
}

Then run in OpenClaw:

/plugin install npm:@fhillipcastillo/agentic-memory
/plugin enable agentic-memory

CLI Commands

| Command | Description | |---------|-------------| | agentic-memory store "content" -t solution | Store a memory | | agentic-memory query "question" | Semantic search | | agentic-memory get "keyword" | Keyword search | | agentic-memory context | Get project context | | agentic-memory patterns | Get solution patterns | | agentic-memory stats | Show statistics | | agentic-memory serve | Start MCP server | | agentic-memory install | Setup MCP |

MCP Tools

When configured, your AI tool has access to:

| Tool | Description | |------|-------------| | memory_store | Store a memory (conversation, solution, conclusion, issue, context) | | memory_get | Search past memories by keyword | | memory_query | Natural language query (requires qmd) | | memory_context | Get project context | | memory_patterns | Get solution patterns | | memory_conclusions | Get conclusions | | memory_issues | Get issues | | memory_stats | Get statistics |

Usage Examples

In AI Conversation

After fixing a bug:
memory_store --content "Fixed auth bug by clearing token on logout" --type solution --tags auth,bug

When making a decision:
memory_store --content "Use qmd for search, MCP for tools" --type conclusion --tags decision

Before starting similar work:
memory_query "how did we handle auth tokens"

Via CLI

# Store a solution
agentic-memory store "Fixed login by clearing localStorage" -t solution -t auth

# Query past solutions
agentic-memory query "how did we solve auth issue"

# Get context
agentic-memory context

# See stats
agentic-memory stats

Configuration

| Env Variable | Default | Description | |--------------|---------|-------------| | MEMORY_DIR | ~/.agentic-memory/ | Storage directory | | MEMORY_PORT | 8181 | MCP server port |

Memory Types

| Type | Description | Storage Dir | |------|-------------|-------------| | conversation | General memories | memories/ | | solution | Solved problems | patterns/ | | conclusion | Decisions | conclusions/ | | issue | Known issues | issues/ | | context | Project context | context/ |

Requirements

• Node.js 18+
• qmd for semantic search: npm install -g @tobilu/qmd

License

MIT