Context that carries forward
Editable long-term memory, session history and reusable skills. Add document retrieval with the optional knowledge base.
PERSONAL AI, SELF-HOSTED
Your agent.
Your workspace.
A personal AI agent that runs on your machine or server. Bring local or cloud models to a built-in WebUI, with memory, tools and automations in one place.
Open sourceApache 2.0Windows · Linux · macOS
Demo could not load. Open the original animation.
BUILT TO CONNECT
Local & cloud modelsModel Context ProtocolMulti-agent tasksScheduled automations
01 / THE WORKSPACE
More than
a chat window.
The pieces a personal agent needs, with a UI to manage them.
Compare your requirementsTools beyond the prompt
Work with files, shell and web tools. Connect external tools through MCP servers managed in the WebUI.
Room to delegate
Run isolated or forked sub-agents for background tasks. Configure project checks to inspect results with real exit codes.
Tasks on a schedule
Set one-shot, interval or cron jobs. Use heartbeat tasks to review a task file and decide whether to act.
Meet your agent where you work
Use the WebUI or connect messaging channels such as Telegram, Discord, Slack and Matrix.
Controls you can inspect
An encrypted credential vault, workspace restrictions, SSRF filtering and prompt-injection defenses in the runtime.
Read the security boundaries02 / YOUR MODELS
Start with the model
you already use.
Connect a local LLM through Ollama or another OpenAI-compatible endpoint, or choose a cloud provider. The integration guides cover setup, authentication and current limitations.
OllamaLocal models
OpenRouterCloud model gateway
OpenAI CodexAccount connection
GitHub CopilotAccount connection
Claude / AnthropicRead access and setup caveats
Provider eligibility, model support, pricing and usage limits apply. Cloud models receive the context sent to them.
Demo could not load. Open the original animation.
03 / YOUR EVERYDAY SETUP
At your desk.
Or on your phone.
The Windows installer includes a desktop launcher. The responsive WebUI also works on your phone when you configure access over your LAN.
Keep your deployment authenticated and protect access beyond localhost. Your model, channel and tool choices determine where data is sent.
Windows installation guideSame agent. Same memory and tools.
A layout that fits the screen.
04 / GET STARTED
Pick your way in.
Install ShibaClaw, connect a model,
then give your agent a task.
DESKTOP
Windows
A bundled desktop app.
No separate Python installation.
irm https://github.com/RikyZ90/ShibaClaw/releases/latest/download/install.ps1 | iexOr paste the command into PowerShell. It downloads and verifies the same setup. Existing ShibaClaw data is preserved.
Installation & updatesAUTO-INSTALLER
Linux / macOS
Install into an isolated Python environment
and start the WebUI with its gateway.
curl -fsSL https://github.com/RikyZ90/ShibaClaw/releases/latest/download/install.sh | bashRequires Python 3.12+. The installer can attempt Python installation on supported systems. Open localhost:3000 after startup.
Installer instructionsCONTAINER
Docker
Run the gateway and WebUI
with the supplied Compose file.
curl -fsSL https://raw.githubusercontent.com/RikyZ90/ShibaClaw/main/docker-compose.yml -o docker-compose.yml
docker compose up -dRun in your chosen installation directory with Docker Compose installed. Open localhost:3000 and create your admin account.
Docker instructionsPYTHON 3.12+
pip
Use your Python environment
on Linux, macOS or Windows.
pip install shibaclaw
shibaclaw web --with-gatewayOpen localhost:3000, create your admin account and follow onboarding.
All installation optionsThe software is open source. Model providers, hosted tools and infrastructure may have their own costs.
05 / KEEP EXPLORING
Open the hood.
Make it yours.
Read the docs, inspect the source,
or find a first contribution.
If ShibaClaw is useful to you, give it a star on GitHub.
It helps other people find the project.