Skip to main content
Velune CLI: a faceted three-dimensional V symbol with five agent nodes (Planner, Coder, Reviewer, Challenger, Synthesizer) converging on it

Velune CLI

Your models. Your machine. Your approval.

An open-source, local-first AI engineer for your terminal.

Visit the website →

PyPI version Python versions CI status License: Apache-2.0 GitHub stars

Status: beta Local-first No analytics MCP client and server Tested on Windows, macOS and Linux


Here's the deal.

AI can write code. Fine. The real question is whether you let it near the codebase that pays your bills.

You've got five worries, and they're all fair:

Five questions teams ask before letting AI touch a codebase, with Velune's answer to each: where your code goes, who checks the answer, what it may change, which models you can use, and whether it knows your project

Velune answers every one of them. And every answer points at a file in this repo. Don't take my word for it. Go check.

What you get

  • 🔒 Your code stays yours. Run a model on your own machine. No analytics, no usage data sent to us (proof).
  • 🧠 A council, not a lone gun. Easy question? One seat. Big refactor? Up to nine seats, including a Challenger whose whole job is to poke holes in the plan.
  • 🛡️ You hold the leash. Manual by default. Every change is a diff you approve. One command to undo.
  • 📚 It remembers. Five memory tiers per project, so it knows your code before you explain it.
A matrix of council seats by depth tier. Instant uses the Coder only, minimal adds the Planner, standard adds the Reviewer and Synthesizer, and full uses all nine seats: five core roles and four specialist critics
A matrix of the Manual, Plan and Auto execution modes against reads, ordinary changes, high-risk changes and changes outside the workspace, showing what is allowed, asked or refused. Manual is the default

It works like this:

/plan add input validation to signup     it investigates and writes a plan. No file touched.
approve                                  it runs the plan. You see every diff first.
/diff                                    see exactly what changed
/rewind                                  don't like it? Undo the agent's last turn.

A council raises your odds of a good answer. It doesn't promise one. It costs more calls than a single model, so it only shows up when the task needs it. Read your diffs.

Plugs into everything

Velune CLI at the centre, connecting to 18 model providers (4 local and 14 cloud), to opt-in application integrations (Docker, HTTP APIs, PostgreSQL, MySQL and MariaDB, Supabase, background processes), to MCP as both server and client, and to plugins, hooks and custom slash commands

The numbers

No hype. Counted from the code. A test recomputes them on every CI run and fails if they drift.

Counted from the repository: 18 provider adapters, 56 slash commands in 12 categories, 5 memory tiers, 9 council seats at full depth, 3 operating systems by 5 Python versions in CI, and 0 analytics services

The honest part: this is beta. Things change between releases. Command execution is process management, not an OS-level sandbox (a Docker sandbox is opt-in). The full list of limits is in SECURITY.md.

Get in. Two minutes.

macOS, Linux, WSL

curl -fsSL https://raw.githubusercontent.com/Surya-Hariharan/Velune-CLI/main/scripts/install.sh | sh

Windows (PowerShell)

irm https://raw.githubusercontent.com/Surya-Hariharan/Velune-CLI/main/scripts/install.ps1 | iex

Or pipx install velune-cli / uv tool install velune-cli. The installer needs no existing Python.

Then pick your lane:

# Fully local
ollama pull qwen2.5-coder:7b
cd your-project && velune init && velune

# Cloud (Groq free tier)
cd your-project && velune init --provider groq && velune setup && velune

Something off? Run velune doctor.

Join in

One maintainer. Open doors. Open an issue · Contribute · Report a vulnerability (privately, never in a public issue).

Go deeper: Reference · Usage guide · Slash commands · Security · Changelog


Licensed under Apache 2.0. Copyright 2026 Surya HA. Third-party names belong to their owners and identify services Velune can connect to. See NOTICE, AUTHORS and the Code of Conduct.

Metadata

Release files for velune-cli 0.9.9

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for velune-cli 0.9.9
File Size Uploaded Provenance Actions
velune_cli-0.9.9.tar.gz 1.7 MB Without provenance Details for velune_cli-0.9.9.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for velune-cli 0.9.9
File Interpreter ABI Platform Actions
velune_cli-0.9.9-py3-none-any.whl Python 3 none any Details for velune_cli-0.9.9-py3-none-any.whl

Total release size: 3.5 MB

Release files / velune_cli-0.9.9.tar.gz

Download URL velune_cli-0.9.9.tar.gz
Size 1.7 MB
Tags Source
SHA-256 checksum
How to use checksums
a1511794edef24d263a24a100b834fc06e6a59110760f6b994950280af18d6c9
BLAKE2b-256 checksum
How to use checksums
5aef6f6425ad01a9e37f84fd7db691b32c5eab237b8a21b922097c96339991ba
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.16

Release files / velune_cli-0.9.9-py3-none-any.whl

Download URL velune_cli-0.9.9-py3-none-any.whl
Size 1.9 MB
Tags Python 3
SHA-256 checksum
How to use checksums
e3c974e81e8f06ee85b02e9293d711b7b8026b3abe9de1be520fbf87efef1c7c
BLAKE2b-256 checksum
How to use checksums
08e3721f8d06fddc63e49022b758efc6914861bd22578a33ac0ec1e14ee4494c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.16

Security for velune-cli 0.9.9

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page