A native macOS AI agent that runs on your Mac. Choose a private local model with Ollama or switch to a cloud model in one click — with voice, memory, skills, scheduled jobs and real tools behind an approval gate.
A fast, native app that stays out of your way — and keeps your data where you decide it belongs.
Run entirely offline with Ollama, or switch to a cloud provider — Groq, OpenCode Zen, DeepSeek, OpenRouter or OpenAI — from the model picker. Your choice, per conversation.
Push-to-talk transcription for your prompts and spoken replies with natural cloud voices via ElevenLabs and Groq Orpheus.
Shama remembers durable facts about you across conversations, so you stop repeating yourself and it keeps getting more useful.
Saved, reusable procedures the agent loads whenever they are relevant — your own playbooks, ready on demand.
Recurring tasks that run in the background via launchd — even while Shama is closed. Set it and forget it.
Reads and edits files and runs shell commands — but always shows a diff preview with an explicit Approve or Reject before anything is written.
Search, rename, delete and export any conversation to Markdown. Your history stays yours and portable.
Hit ⌘K to switch model, language, skill or session. Five languages: English, Haitian Creole, French, Spanish and Portuguese.
Connects to a local OpenCode server and borrows its agents, MCP tools and 200+ providers — extending Shama without changing how you work.
Shama runs natively on your Mac. Nothing leaves your machine unless you choose a cloud model — the choice is yours, per conversation.
Grab the latest build from GitHub Releases — no store, no account, no telemetry.
Native macOS app · Intel (x64) build · macOS 13 or later
Download from GitHub Releases