aimem-cli 0.2.6

AI Memory Switcher — Save, compress, and transfer context between AI agents

🧠 AiMem — AI Memory Switcher

Transfer context seamlessly between AI agents (Claude, Gemini, Qwen, OpenCode, Codex...)

When Claude hits rate-limits or you need to switch to another AI, AiMem helps you continue without losing your train of thought.

If this project helps you, please give it a ⭐ on GitHub!

---

🤔 The Problem

You: "Claude is running out of tokens..."
─────────────────────────────────────────────────
│ Claude: Fixing authentication bug...            │
│ 11PM - Claude hits rate limit!                │
│                                              │
│ NOW WHAT?                                    │
│ → Copy all chat history? Tedious             │
│ → Start from scratch? 30+ minutes lost      │
│ → Lose context? Devastating                 │
─────────────────────────────────────────────────

→ With AiMem: 3 seconds to transfer
$ aimem save --from claude
$ aimem load sess-abc --to gemini
# Paste into Gemini - full context transferred!

---

🚀 Install

pip install aimem-cli

Or from source:

git clone https://github.com/ThangTo/AiMem.git
cd AiMem
pip install -e .

---

✨ Features

• 🎮 Interactive TUI: Fully featured Terminal UI with arrow-key navigation
• 📤 Save sessions from Claude, Gemini, Qwen, OpenCode, Codex, Aider, Continue.dev
• 📥 Load to any AI agent with proper format
• 💉 Inject directly into agent storage (auto-opens new terminal to resume!)
• 📊 Analyze if session fits target model context limit
• ✂️ Chunk large sessions automatically
• 🗜️ Compress sessions with LLM (95% reduction, configurable via TUI settings)
• 🔀 Merge multiple sessions (Smart deduplication)
• 🗑️ Multi-Delete: Select and delete multiple sessions easily via TUI

---

⚡ Quick Start

🎮 Interactive Mode (TUI - Recommended)

The easiest way to use AiMem is via the interactive Terminal UI. Just run:

aimem-cli

Use arrow keys to navigate menus, configure settings, compress sessions, and transfer context without remembering any commands!

---

Transfer from Claude to OpenCode (CLI)

# Save current Claude session
aimem-cli save --from claude

# Inject directly into OpenCode
aimem-cli load <session-id> --to opencode --inject

# Open the session in OpenCode
opencode -s <session-id-from-output>

Transfer to Markdown (for any AI)

aimem-cli save --from claude
aimem-cli load <session-id> --to markdown
# Content copied to clipboard - paste anywhere

---

📖 All Commands

aimem-cli init

Initialize config file at ~/.aimem/config.json.

aimem-cli init
aimem-cli init --force    # Overwrite existing config

---

aimem-cli save --from <agent>

Save a session from an agent to AiMem storage.

# From Claude
aimem-cli save --from claude
aimem-cli save --from claude --session-id <id>

# From Gemini
aimem-cli save --from gemini

# From Qwen
aimem-cli save --from qwen

# From OpenCode
aimem-cli save --from opencode

# From Codex
aimem-cli save --from codex

# From Aider
aimem-cli save --from aider

# From Continue.dev
aimem-cli save --from continue

# From Clipboard
aimem-cli save --from clipboard

# With LLM compression (requires API key)
aimem-cli save --from claude --compress

---

aimem-cli load <session-id> --to <format>

Load a saved session to target format.

# To Markdown (copy to clipboard)
aimem-cli load <session-id> --to markdown

# To Claude (markdown format)
aimem-cli load <session-id> --to claude

# To Gemini
aimem-cli load <session-id> --to gemini

# To Qwen
aimem-cli load <session-id> --to qwen

# To Prompt (for API calls)
aimem-cli load <session-id> --to prompt

# To OpenCode
aimem-cli load <session-id> --to opencode

# To Codex
aimem-cli load <session-id> --to codex

# To Continue.dev
aimem-cli load <session-id> --to continue

---

aimem-cli load <session-id> --to <agent> --inject

Inject directly into agent storage - no copy-paste needed!

# Inject into OpenCode (RECOMMENDED)
aimem-cli load <session-id> --to opencode --inject
opencode -s <session-id>

# Inject into Claude
aimem-cli load <session-id> --to claude --inject
claude --resume

# Inject into Gemini
aimem-cli load <session-id> --to gemini --inject

# Inject into Qwen
aimem-cli load <session-id> --to qwen --inject

# Inject into Codex
aimem-cli load <session-id> --to codex --inject

---

aimem-cli list

List all saved sessions.

# List all saved sessions
aimem-cli list

# List sessions from specific agent
aimem-cli list --from claude
aimem-cli list --from gemini
aimem-cli list --from qwen

# List available agents
aimem-cli list --agents

---

aimem-cli merge <session-id-1> <session-id-2>

Merge multiple sessions into one.

# Simple merge (concatenate)
aimem-cli merge sess-abc sess-def

# Smart merge (deduplicate, combine goals, merge todos)
aimem-cli merge sess-abc sess-def --smart

# Smart merge with target format
aimem-cli merge sess-abc sess-def --smart --to gemini

---

aimem-cli config

View or update configuration.

# Show current config
aimem-cli config

# Set config values
aimem-cli config set compression.enabled true
aimem-cli config set compression.api_key YOUR_API_KEY
aimem-cli config set compression.provider groq
aimem-cli config set output.format markdown
aimem-cli config set output.clipboard_auto true

---

aimem-cli delete <session-id>

Delete a saved session.

aimem-cli delete <session-id>

---

Analyze context before loading

Check if a session fits in target model's context limit.

# Analyze if session fits in Qwen's 32K context
aimem-cli load <session-id> --to qwen --analyze

# Output example:
# ⚠️ Session EXCEEDS Qwen limit (50K tokens vs 32K limit)
# 💡 Try: aimem load <session-id> --to qwen --chunk

---

Chunk large sessions

Automatically split session into smaller chunks if too large for target.

# Auto-chunk for Qwen (32K limit)
aimem-cli load <session-id> --to qwen --chunk

# Output:
# 📦 CHUNK 1/2 (25K tokens)
# [content of chunk 1]
# ─────────────────────
# 📦 CHUNK 2/2 (25K tokens)
# [content of chunk 2]

---

Compress session with LLM

Reduce session size using LLM compression (requires API key).

# Compress on save
aimem-cli save --from claude --compress

# Compress on load
aimem-cli load <session-id> --compress

# Result: ~95% size reduction
# Output: {
#   "current_goal": "Fix authentication bug",
#   "latest_code": [{"path": "auth.py", "content": "..."}],
#   "current_errors": ["TypeError: undefined"],
#   "key_decisions": ["Used JWT instead of session"],
#   "todo_list": ["Write tests", "Deploy to staging"]
# }

---

🔧 Supported Agents

Source (--from)

Agent	Storage Location	Inject?
claude	~/.claude/projects/*/*.jsonl	✅
gemini	~/.gemini/tmp/*/chats/*.json	✅
qwen	~/.qwen/tmp/*/logs.json	✅
opencode	~/.opencode/sessions/*.json	✅
codex	~/.codex/sessions/*.jsonl	✅
aider	~/.aider.chat.history.md	❌
continue	~/.continue/sessions.db	❌
clipboard	System clipboard	❌

Target (--to)

Format	Best for	Inject?
markdown	Paste anywhere	❌
claude	Claude Code CLI	✅
gemini	Gemini CLI	✅
qwen	Qwen CLI	✅
opencode	OpenCode CLI	✅
codex	Codex CLI	✅
prompt	API calls	❌

---

⚙️ Configuration

Config file: ~/.aimem/config.json

# Show config
aimem-cli config

# Enable LLM compression (opt-in)
aimem-cli config set compression.enabled true
aimem-cli config set compression.api_key YOUR_GROQ_KEY
aimem-cli config set compression.provider groq

# Auto-copy to clipboard
aimem-cli config set output.clipboard_auto true

Config Options

Option	Default	Description
compression.enabled	false	Enable LLM compression
compression.provider	"groq"	Compression provider (groq, gemini)
compression.api_key	""	API key for compression
output.format	"markdown"	Default output format
output.clipboard_auto	true	Auto-copy to clipboard
storage.path	"~/.aimem/sessions"	Session storage path

---

📝 Examples

Example 1: Transfer Claude session to OpenCode

$ aimem-cli save --from claude
[i] Found 5 Claude sessions. Select one:
  [1] 2026-04-20 | Fix authentication bug
  [2] 2026-04-19 | Add user login feature
  [3] 2026-04-18 | Research API design
Enter number (default=1): 1
[OK] Saved session: claude-abc123

$ aimem-cli load claude-abc123 --to opencode --inject
[OK] Injected into OpenCode
    Session ID: ses_xyz789
Resume with: opencode -s ses_xyz789

$ opencode -s ses_xyz789
# Continue with full context in OpenCode!

Example 2: Transfer to any AI via Markdown

$ aimem-cli save --from claude
[OK] Saved session: claude-abc123

$ aimem-cli load claude-abc123 --to markdown
[OK] Content copied to clipboard!

# Paste into any AI: Gemini, ChatGPT, Qwen, etc.

Example 3: Merge multiple sessions

$ aimem-cli merge claude-abc claude-def claude-ghi
[i] Merging 3 sessions...
[OK] Merged session: merged-xyz123
    Total messages: 156

$ aimem-cli load merged-xyz123 --to gemini --inject

---

🏗️ Architecture

aimem-cli/
├── aimem/
│   ├── cli.py                 # CLI interface
│   ├── models.py              # UniversalSession model
│   ├── storage.py             # File storage
│   ├── compression.py         # LLM compression
│   ├── context_manager.py     # Context management
│   └── adapters/
│       ├── claude.py          # Claude Code adapter
│       ├── gemini.py          # Gemini CLI adapter
│       ├── qwen.py             # Qwen CLI adapter
│       ├── opencode.py         # OpenCode adapter
│       ├── codex.py            # Codex adapter
│       ├── aider.py            # Aider adapter
│       ├── continue_dev.py     # Continue.dev adapter
│       ├── clipboard.py        # Clipboard adapter
│       └── output/             # Output formatters
│           ├── __init__.py
│           ├── markdown.py
│           ├── claude.py
│           ├── gemini.py
│           ├── qwen.py
│           ├── prompt.py
│           ├── continue.py
│           ├── codex.py
│           └── opencode.py

---

---

📋 Supported Agents

Source	Storage	Inject?
Claude	~/.claude/projects/*/*.jsonl	✅
Gemini	~/.gemini/tmp/*/chats/*.json	✅
Qwen	~/.qwen/tmp/*/logs.json	✅
OpenCode	~/.opencode/sessions/*.json	✅
Codex	~/.codex/sessions/*.jsonl	✅
Aider	~/.aider.chat.history.md	❌
Continue	~/.continue/sessions.db	❌
Clipboard	System clipboard	❌

---

🚀 Roadmap

• Add more adapters ( Zed, Cursor )
• Docker support
• WebUI for session management
• VS Code extension

---

📝 License

MIT License - See LICENSE

---

👏 Contributing

See CONTRIBUTING.md for how to contribute.

---

👏 Acknowledgments

• Claude Code for the great developer experience
• All open source contributors

---

_{Made with ❤️ by ThangTo}