its-over-9k - The AI Memory Framework

Your AI forgets everything between sessions. its-over-9k fixes that — and a lot more.

One load_project() call. ~3000 tokens. Your agent knows everything important about a project — every past mistake, every decision, every open task — across sessions, devices, and AI providers. No setup per conversation. No "let me re-read the codebase." It just remembers.

AI agent? Skip this file. Read AGENT_SETUP.md — written for you, not for humans.

What This Is

its-over-9k is not a note-taking plugin. It's a memory framework for AI agents — a complete infrastructure layer for persistent, portable, token-efficient knowledge that survives session boundaries, device switches, and provider changes.

Four core guarantees:

The Problem

Every AI session starts from zero. Your agent asks the same questions, makes the same mistakes, contradicts last week's decisions, and wastes 50k tokens loading context it already processed yesterday.

You've tried workarounds — CLAUDE.md files, custom prompts, manually pasting context. They don't scale. You have 10 projects. You switch between 3 devices. You use different AI tools.

The Solution

You:    "Load project"
Agent:  [calls load_project("P0048") — 3000 tokens]
Agent:  "v1.2.9, TypeScript/SQLite/npm. 3 open bugs, 8 roadmap items.
         Last session: rebrand complete, rename_id bug fixed (89 changes).
         Next: O-Entry Auto-Purge. What's the focus today?"

That's it. 3000 tokens for a complete project briefing. The agent knows the stack, the architecture, the open bugs, the recent decisions, and exactly where you left off — even if "you" was a different AI on a different machine yesterday.

How It Works

Level 1  ──  One-line summary          (always loaded — ~5k tokens for 300+ entries)
  Level 2  ──  Paragraph detail        (loaded on demand)
     Level 3  ──  Full context          (loaded on demand)
      Level 4  ──  Extended detail      (loaded on demand)
        Level 5  ──  Raw/verbatim data  (loaded on demand)

At session start, the agent loads Level 1 summaries — one line per memory. When it needs detail, it drills down. Your 300-entry memory costs 5k tokens to overview. A single project costs ~3000 tokens.

Nothing is summarized away. Level 1 is a compressed view, but Levels 2–5 hold the complete original text, word for word, accessible on demand.

Framework Features

Automatic Session Memory

Every conversation is recorded automatically. No "save your work" prompts. No manual checkpoints.

You type  →  Agent responds  →  Stop hook fires  →  Exchange saved to O-entry
                                                   →  Linked to active project
                                                   →  Haiku auto-titles the session

Switch projects mid-session? The O-entry switches too. Start a new session on a different device? The next agent sees every exchange from every device — the conversation never dies.

Haiku Background Checkpoints

Every N exchanges (configurable, default 5), a Haiku subagent wakes up in the background. It reads the recent conversation, extracts lessons learned, errors encountered, and decisions made, then writes them to long-term memory — with full MCP tool access. Your main agent is never interrupted.

The checkpoint also writes a handoff note to the project: "Here's what was done, here's what's in progress, here's the next step." The next agent — on any device, any provider — picks up exactly where you left off.

Project-Based, Not Session-Based

Sessions are meaningless. Projects are everything.

• O-entries are linked to the active project, not the session
• Checkpoint counters count project exchanges, not session messages
• load_project shows recent conversations with full context — across all devices

Skills System

its-over-9k ships with a complete skills layer — structured behavior files that agents load on demand. Skills define how an agent should do something (debug, write memory, curate entries, handle a session start) — separate from memory, separate from prompts.

npx hmem update-skills    # Pull latest skills to your AI tool's skill directory

Skills are versioned and updated independently. Your agents get smarter without reinstalling. 21 skills ship by default:

Company Memory

Beyond personal memory, agents can maintain a shared company store — a separate company.hmem that multiple agents and team members can read from. Personal and company memory coexist; agents query both simultaneously.

import { openCompanyMemory } from 'its-over-9k';
const store = openCompanyMemory('/path/to/project');

Embeddable SDK

its-over-9k ships as a fully documented TypeScript SDK — import HmemStore directly into your own agents, tools, or automation pipelines:

import {
  HmemStore, openCompanyMemory, resolveHmemPath,
  loadHmemConfig, saveHmemConfig, DEFAULT_CONFIG, DEFAULT_PREFIXES, formatPrefixList,
  searchMemory,
} from 'its-over-9k';
import type {
  AgentRole, MemoryEntry, MemoryNode, HmemConfig,
  SearchResult, SearchOptions, SearchScope,
} from 'its-over-9k';

const store = new HmemStore('/path/to/agent.hmem');
const results = searchMemory('/path/to/project', 'auth token bug', { maxResults: 5 });

MCP Tools

its-over-9k ships two MCP servers:

• hmem (daily-use, 13 tools) — read, write, search, project lifecycle.
• hmem-curate (maintenance, 11 tools) — bulk edits, schema migrations, backup/restore. Activate only when curating.

hmem — daily-use server (13)

hmem-curate — maintenance server (11)

The daily hmem server is registered automatically by hmem init. hmem-curate is opt-in because its tools are destructive — register it when you need to curate:

npx hmem mcp-add-curate          # adds the curate server to Claude Code (user scope)

Then toggle it via /mcp only during curation sessions. For other MCP clients, see Manual setup.

CLI Commands

After npm install -g its-over-9k, the hmem binary is on PATH.

User-facing

Hook drivers (called by AI tools, not by hand)

Curation

Sync (requires hmem-sync installed)

Backup / migration

hmem serve starts the MCP stdio server directly — your AI tool launches it automatically; you only run it by hand for debugging.

Memory Categories

Default prefixes (configurable via prefixes in hmem.config.json):

Add custom prefixes (e.g. A for App, F for Function reference) by listing them under prefixes in hmem.config.json — they show up in read_memory({ prefix: "X" }) filters automatically.

Quick Start

1. Install

npm install -g its-over-9k

2. Run the interactive installer

npx hmem init

Detects your AI tools, creates the memory directory, configures MCP, and installs all hooks:

3. Verify

Restart your AI tool, then:

read_memory()

Empty response = working (first run). Error = check the troubleshooting section.

Manual setup

{
  "mcpServers": {
    "hmem": {
      "command": "/absolute/path/to/node",
      "args": ["/absolute/path/to/its-over-9k/dist/mcp-server.js"],
      "env": {
        "HMEM_PROJECT_DIR": "/home/yourname/.hmem",
        "HMEM_AGENT_ID": "DEVELOPER"
      }
    }
  }
}

echo "Node: $(which node)"
echo "Server: $(npm root -g)/its-over-9k/dist/mcp-server.js"

{
  "mcp": {
    "hmem": {
      "type": "local",
      "command": ["/absolute/path/to/node", "/absolute/path/to/its-over-9k/dist/mcp-server.js"],
      "environment": { "HMEM_PROJECT_DIR": "/home/yourname/.hmem" },
      "enabled": true
    }
  }
}

{
  "mcpServers": {
    "hmem": {
      "command": "/absolute/path/to/node",
      "args": ["/absolute/path/to/its-over-9k/dist/mcp-server.js"],
      "env": { "HMEM_PROJECT_DIR": "/home/yourname/.hmem" }
    }
  }
}

Configuration

hmem.config.json in your HMEM_PROJECT_DIR (or Agents/NAME/):

{
  "memory": {
    "maxCharsPerLevel": [200, 2500, 10000, 25000, 50000],
    "maxDepth": 5,
    "checkpointMode": "auto",
    "checkpointInterval": 5,
    "recentOEntries": 10,
    "maxTitleChars": 50,
    "prefixes": { "X": "Custom" }
  },
  "sync": {
    "serverUrl": "https://your-server/hmem-sync",
    "userId": "yourname",
    "salt": "...",
    "token": "..."
  }
}

All keys are optional. Missing keys use defaults.

Checkpoint setup per harness

The auto-checkpoint agent runs in the background after every Nth exchange. It needs an LLM call — three paths, picked automatically:

1. API key in environment (any harness) → direct provider API loop. Configure checkpointProvider + checkpointModel + checkpointApiKeyEnv in hmem.config.json. Works from Pi, Hermes, OpenCode, and Claude Code.
2. No API key, but claude CLI in PATH → subprocess fallback (claude -p). Zero-config for Claude Code / Claude Max users.
3. Neither → checkpoint fails with a config-hint error.

Recommended cheap setup (DeepSeek, ~10× cheaper than Haiku):

{
  "memory": {
    "checkpointMode": "auto",
    "checkpointProvider": "openai",
    "checkpointModel": "deepseek-chat",
    "checkpointBaseUrl": "https://api.deepseek.com/v1",
    "checkpointApiKeyEnv": "DEEPSEEK_API_KEY"
  }
}

Then export DEEPSEEK_API_KEY=sk-... in your shell profile. Works for any harness.

Claude Code / Claude Max (zero-config): no provider settings needed — the subprocess fallback uses your existing claude login.

Per-harness exchange logging: Claude Code uses Stop hooks (installed by npx hmem init). Pi uses the built-in extension (src/extensions/pi-hmem.ts). Hermes needs the hermes-hmem plugin (see plugins/hermes-hmem/README.md). OpenCode uses the same hook system as Claude Code.

⚠ Pi users: Hook-based exchange logging is currently not working in the Pi runtime. Exchanges are not saved to O-entries automatically. Manual checkpointing via hmem checkpoint is the workaround until this is resolved.

Cross-Device Sync

Sync memories across all devices with zero-knowledge AES-256-GCM encryption.

npm install -g hmem-sync
npx hmem-sync connect     # Interactive wizard — first device creates, others join

Add HMEM_SYNC_PASSPHRASE to your MCP config for automatic sync on every read/write.

Multi-server redundancy

{
  "sync": [
    { "name": "primary", "serverUrl": "https://server1/hmem-sync", "userId": "me", "salt": "...", "token": "..." },
    { "name": "backup",  "serverUrl": "https://server2/hmem-sync", "userId": "me", "salt": "...", "token": "..." }
  ]
}

Announcements

Broadcast to all synced agents across all devices:

npx hmem-sync announce --message "Server URL changing — update your config!"

Windows

On Windows with Git for Windows, Claude Code routes hook and statusLine commands through Git Bash by default. Git Bash's MSYS2 runtime crashes transiently at startup, killing the command before it runs.

Fix: add "shell": "powershell" to every hook command and to statusLine in ~/.claude/settings.json.

See settings.windows.example.json for the full working config. Key differences:

{
  "env": {
    "HMEM_PATH": "C:/Users/YOUR_USERNAME/.hmem/Agents/DEVELOPER/DEVELOPER.hmem"
  },
  "hooks": {
    "Stop": [{
      "matcher": "",
      "hooks": [{
        "type": "command",
        "command": "node C:/Users/YOUR_USERNAME/AppData/Roaming/npm/node_modules/its-over-9k/dist/cli.js log-exchange",
        "shell": "powershell"
      }]
    }]
  },
  "statusLine": {
    "type": "command",
    "command": "node C:/Users/YOUR_USERNAME/AppData/Roaming/npm/node_modules/its-over-9k/dist/cli.js statusline",
    "shell": "powershell"
  }
}

Run npm root -g to get the correct node_modules path for your machine.

statusLine on Windows: Stable with "shell": "powershell". Without it the statusline disappears intermittently.

Troubleshooting

Updating

npm update -g its-over-9k    # MCP server + SDK
npm update -g hmem-sync      # Sync (if installed)
npx hmem update-skills       # Refresh skill files

License

MIT