AlysisAi/alysis-code: Everyone's coding agent

Local CLI coding agent that turns plans into reviewed, PR-ready code.
Bring your own model. Sandboxed by default.

Website · Docs · Changelog

Get Started

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

1 — Install

pipx install alysis-code

2 — Launch in your project

cd your-project
alysis-code

The first run opens a guided setup wizard: choose Use an API key or Use an AI subscription, connect a provider, and pick your default model. Re-run it anytime with alysis-code setup. (alysis works too, as a short alias.)

3 — Start building

alysis-code chat                            # interactive coding session
alysis-code run "Fix the failing tests"     # one-shot task

Bring your own key from any supported provider — or point Alysis Code at a local endpoint (Ollama, LM Studio, vLLM) and run fully offline. The /config picker lists every preset. Full walkthrough: Quickstart guide.

Why Alysis Code

• Easy to use — a full coding agent behind a clean, convenient terminal interface that anyone can

• Built for every developer — from solo scripts to team servers and CI, the same agent fits every

• Forge Mode — plan-driven execution for larger work:

• Orchestrated subagents — isolated workers run in

• Free built-in web search — model-independent web_search that works with any provider, at no

• Bring your own model — OpenAI, Anthropic, DeepSeek,

• Deeply extensible — skills, custom tools, MCP

• Sandboxed by default — Docker or Bubblewrap

Shell and verification commands default to strict sandboxing. Verification can run outside that sandbox only when explicitly configured with verify_sandbox.mode="off" or ALYSIS_VERIFY_SANDBOX_MODE=off.

Modes & Personas

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

Execution modes are the enforcement layer. Choose one per command with --mode, select the next message's permissions with /permissions or Shift+Tab, or set a default with alysis-code config set default_mode . A selection leaves the active turn's permissions unchanged until the next message starts:

| Mode | Behavior | | --- | --- | | readonly | Inspection only — no file writes, shell, MCP, or subagent delegation. | | review | The safe default — previews and asks before every write and shell command. | | auto | Applies routine edits with fewer prompts; dangerous operations stay blocked. | | fullaccess | No mode-level prompts for trusted workspaces — the denylist and audit log stay active. |

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

| Persona | Posture | | --- | --- | | code | Implementation — the default, full coding workflow. | | architect | Planning and design — may write markdown documents only. | | ask | Read-only questions and codebase exploration. | | debug | Reproduce the failure first, then fix minimally and rerun the reproduction. |

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 legacy chat Plan Mode and its /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.

FAQ

More questions? See the full FAQ.

Join the community — X | GitHub Issues