alysis-code 0.16.0

An open-source agent for coding, research, and everyday workflows.
Multiple models. Our own sandbox. One independently built engine.

Website · Documentation · Quickstart · Changelog

Follow on X · Sponsor the project

_{From the welcome screen to your first conversation. Real CLI recording; pauses shortened.}

Get Started

Alysis Code requires Python 3.11+ and runs on macOS, Linux, and Windows.

1 — Install

On Windows, paste this into PowerShell:

irm https://raw.githubusercontent.com/AlysisAi/AlysisCode/main/scripts/install.ps1 | iex

On Linux, macOS, or inside a WSL/Ubuntu terminal:

curl -LsSf https://raw.githubusercontent.com/AlysisAi/AlysisCode/main/scripts/install.sh | sh

The installers provision uv and Python 3.12 when needed. No existing Python or pipx is required. PowerShell can use alysis immediately; after the shell installer, follow its PATH instructions or open a new terminal. If you already use pipx, pipx install alysis-code also works; run pipx ensurepath and restart your terminal if the command is missing. Already installed? Run alysis update to get the latest release.

2 — Choose how to connect to a model

Pick one path. You can switch later with /config or by re-running alysis-code setup.

Path A — Your own API key or a local model (no Alysis account needed)

cd your-project
alysis-code

The first run opens a guided setup wizard. Choose Use an API key, connect a provider (OpenAI, Anthropic, DeepSeek, Gemini, OpenRouter, and more), and pick your default model. To run fully offline, point Alysis Code at a local endpoint (Ollama, LM Studio, vLLM) from the /config picker, which lists every preset. Prefer to configure it by hand?

export ALYSIS_API_KEY="YOUR_KEY"
alysis config set base_url "https://api.openai.com/v1"
alysis config set model "gpt-4.1-mini"

Have a ChatGPT subscription instead? Choose Use an AI subscription in the wizard, or run alysis auth login openai-codex.

Path B — Alysis account

alysis login

Approve the code in your browser to use your Alysis credits. No provider key and no cloud environment variables are needed.

3 — Start working

cd your-project
alysis-code chat                            # interactive agent session
alysis-code run "Fix the failing tests"     # coding task
alysis-code run "Summarize the notes in this folder into a project brief"

(alysis works too, as a short alias.) The /config picker lists every provider preset. Full walkthrough: Quickstart guide.

Alysis Code App

Alysis Code App is powered by the open-source Alysis Code engine, bringing an agentic desktop experience to coding and broader everyday work.

Why Alysis Code

Alysis Code combines our own sandbox, multiple model providers, and general-purpose workflows in an independently built, Apache-2.0 agent. Build software, research a question, prepare documents, or automate work from the same terminal.

• Our own Alysis Sandbox — command execution is sandboxed by default, with Alysis-maintained container images and native OS backends. Permissions, network restrictions, and repository protection give you concrete controls over what the agent can do.
• Multi-model by design — use OpenAI, Anthropic, Gemini, DeepSeek, Qwen, Mistral, OpenRouter, xAI, and more. Connect your own key without an Alysis account, or use Ollama, LM Studio, or vLLM for local inference. Choose the model that fits the task.
• General workflows, too — research with built-in web search, turn notes into briefs, and automate tasks across your files. Optional document tools add PDF, Word, spreadsheet, and presentation workflows. Web search needs no separate API key; model usage follows your provider's pricing.
• An independently built agent engine — our runtime, Forge orchestration, subagent system, terminal interface, and sandbox layer are developed by Alysis AI. The open-source CLI lets you inspect, extend, and run the engine yourself.
• Forge: turn a plan into verified changes — execute tasks in dependency order, retain verification and review evidence, and resume interrupted runs. Strict verification blocks integration when checks fail; independent tasks can run in parallel with forge swarm.
• Parallel work you can inspect before applying — isolated workers prepare changes in separate Git worktrees. A verifier can check the exact candidate while your working tree stays unchanged; applying it runs host-side conflict checks.
• Extend the whole workflow — package skills, custom tools, MCP servers, hooks, and specialist subagents together in a declarative plugin manifest.

Alysis Sandbox

Our agent. Our sandbox. Your control. Alysis Sandbox is our own open-source execution layer, with Alysis-maintained container images and Docker, Linux Bubblewrap, or macOS Seatbelt backends. Shell and verification commands default to strict sandboxing. The runtime enforces these boundaries outside the model, so command protection does not depend on the model following a safety prompt.

• Controlled execution. Networking defaults to off; the macOS backend still permits localhost. Docker commands drop Linux capabilities and do not inherit the host environment by default. The CLI refuses execution when no backend is available unless you explicitly configure a fallback.
• Our own development images. ghcr.io/alysisai/alysis-sandbox provides base, dev, and server variants. The image build uses pinned, verified inputs for both AMD64 and ARM64; deployments can pin an image digest for repeatable environments.
• Protection beyond approval prompts. Repository metadata protection and the command denylist remain separate from permission modes, including fullaccess.

Set up and check your sandbox:

alysis sandbox setup
alysis sandbox doctor --smoke

The sandbox covers shell and verification execution; the workspace remains writable for the task. See the sandbox guide for backend differences and setup, and the image build for the supply-chain details.

How it compares

Comparing current terminal agents: Claude Code, Codex CLI, OpenCode, Command Code, and Kilo Code. Alysis Code brings together our own sandbox with strict defaults, multiple providers and local models, and coding and general workflows. For larger coding tasks, Forge adds coordinated execution, isolated worktrees, and verification before integration — in an independently built, Apache-2.0 engine.

Based on the linked official documentation, checked October 25, 2026. This compares the CLI workflows, not every desktop or cloud feature. Capabilities and defaults vary by platform and configuration; the table is not a performance benchmark.

Permissions & Personas

Two layers control what the agent may do — one enforces, one focuses.

Permissions are the enforcement layer. Choose one per command with --mode, select the next message's level in chat with /permissions or Shift+Tab, or set a default with alysis-code config set default_mode <mode>. A selection takes effect when the next user message starts; it does not change a turn already running:

Personas are working postures layered on top. Switch with /persona, or cycle with Tab in the TUI:

A persona is a convention, never a permission: it can keep or lower your execution mode, but can never raise it. Define your own in .alysis_personas/*.md. Learn more

Use /persona architect for planning in chat, or /forge for a plan-driven execution workflow. The former chat /plan commands have been removed; Forge retains its plan views and editor. Persona changes made during a running turn apply after that turn finishes.

Project Links

• Website
• Documentation — also mirrored in docs/
• Changelog
• Contributing — local setup and PR expectations
• Security policy — report vulnerabilities privately, never through public issues

License

Apache-2.0. You're free to use, modify, and distribute this code, including commercially, as long as you keep the attribution and license notices. See LICENSE and NOTICE. The shared editor sidebar includes MIT-licensed adaptations from Kilo Code/OpenCode; retain its third-party notices.

FAQ

More questions? See the full FAQ.

Join the community — X | GitHub Issues

Editor integrations

VS Code uses the shared Alysis sidebar and Python IDE bridge. JetBrains is a development preview with a smaller feature set. The VS Code Marketplace pre-release is pending production signing and acceptance; see the IDE release status before installing component packages.