SuitcaseYour private AI assistant that learns and grows with you.

One command to start. Bring your own LLM. Your memory, tools, and dreams stay local.

Quickstart
View on GitHub
🧳

A Companion, Not Just a Chat Box

Suitcase keeps a living model of your goals, evidence, preferences, and operating style so it can get more useful the longer you work with it.

💤

Dreams Into Better Memory

Short-term memory captures recent work. Dream cycles garbage-collect the noise, distill durable patterns, and preserve what should shape future help.

🤖

LLM-Native

Defaults to OpenAI with gpt-4o, task-scoped models (gpt-4o-mini for fast tasks), and vision support. Also supports OpenRouter, Anthropic, and any OpenAI-compatible endpoint.

⚡

27 Powers

Curl, jq, ripgrep, git, GitHub CLI, ffmpeg, pandoc, tesseract, rclone, restic, Discord, Telegram, Brave Search, and more — toggle from the setup dashboard.

🧠

Agent With Bash

The Synax agent has native bash access. SKILL.md files provide guidance. No Python middleware between the agent and its tools.

💬

Multi-Channel Memory

Web chat, Discord DMs, or Telegram. Same assistant, same evolving memory, across all channels.

🏠

Local-First

No cloud dependency. Career data, profile, memories, reflections, and conversation history live in Docker volumes on your machine.

🔄

Self-Updating

Suitcase checks for its own updates. Agent says "I'll be right back" — you run ./scripts/update.sh to pull, rebuild, and restart.

How It Works ​

You
web chat
Web UI
DMs / Telegram
Discord / Telegram
Suitcase API
FastAPI + Docker
inference
Your LLM
state + memory
Docker Volumes
/data /world /logs

Suitcase is a local-first AI assistant in Docker. The Synax agent has native bash access to installed CLIs, while Suitcase keeps short-term and long-term memory so it can learn how you work. Its dream cycles turn raw interaction history into compact, durable self/user/world updates instead of letting memory become an infinite junk drawer.

Memory That Grows With You ​

Suitcase treats memory as part of the product, not an afterthought:

  • Short-term memory captures recent conversations, tool use, reflections, and working context.
  • Dream cycles periodically review that short-term memory, garbage-collect low-value noise, and synthesize useful long-term representations.
  • Long-term memory stores durable facts, preferences, strategies, and self-model updates that help the assistant become more aligned over time.
  • World files (WORLD.md, SOUL.md, SELF.md, and USER.md) give those memories a bounded home with explicit review rules.

The goal is not a bigger chatbot transcript. The goal is an assistant that notices patterns, improves its model of you, and becomes less uselessly generic every week.

Ready in One Command ​

bash
curl -fsSL https://raw.githubusercontent.com/achuthanmukundan00/Suitcase/main/scripts/install.sh | bash

Or if you prefer manual control:

bash
git clone https://github.com/achuthanmukundan00/Suitcase.git
cd Suitcase
cp .env.example .env   # set LLM_API_KEY
./start.sh             # builds frontend, starts Docker

Open http://localhost:8000/setup — connect your LLM, enable the powers you want, and start chatting.

Your private career copilot.