Run a local AI agent desktop app without depending on a closed official app.
Set up a self hosted shared workspace so a team can continue each other's AI agent sessions.
Connect the agent to a self hosted or OpenAI compatible model provider through the Chat Completions bridge.
Use locally installed skills and plugins to generate documents, spreadsheets, or presentations from the desktop app.
| hellanglez/zspark | geekgineer/needle-rs | ipetkov/conch-runtime | |
|---|---|---|---|
| Stars | 27 | 26 | 26 |
| Language | Rust | Rust | Rust |
| Last pushed | — | — | 2021-05-24 |
| Maintenance | — | — | Dormant |
| Setup difficulty | hard | moderate | moderate |
| Complexity | 4/5 | 4/5 | 2/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires building the Rust Codex runtime plus Node.js, and Docker for the optional shared server.
zspark is an open source desktop and collaboration shell built on top of the open source Codex runtime, the same engine used by tools like Codex App. It is not affiliated with OpenAI, but it forks and reuses the open Codex runtime as its agent engine. The idea is to take product features that normally live inside closed desktop apps, such as a local agent desktop, shared team sessions, enterprise login, downloadable file outputs, and plugin support, and make them available in a self hostable project instead. The project gives you a desktop agent app built with Electron that wraps the Codex runtime, so you get local AI sessions, memory, tool use, and permission controls without needing the official closed Codex App. It can discover and use locally installed Codex skills and plugins, covering things like generating presentations, documents, and spreadsheets when the right tools are present. You can attach files to a conversation and the agent can hand back generated files as downloadable artifacts inside the chat. On the model side, zspark can talk directly to providers using OpenAI's Responses API, and it also includes a bridge for providers that only support the older Chat Completions format, such as many OpenAI compatible gateways and self hosted model servers. For teams, there is an optional self hosted server that stores shared workspaces, sessions, and artifacts so multiple people can pick up each other's work. It uses Postgres for storage and Redis together with real time collaboration tools to keep everyone in sync. Microsoft Entra ID is built in for controlling who can access these shared workspaces, with support aimed at both standard setups and Azure China. The project ships desktop builds for macOS and Windows, while the shared server is meant to run on Linux. Setup requires Git, Node.js 22 or newer, Rust, and Docker if you want to run the shared server. The README describes this as an early release: the core desktop loop, provider setup, the Chat Completions bridge, the shared server, and Entra based login all work, but the project is still moving fast and some parts may be rough.
An open source, self hostable desktop and team collaboration app built around the Codex AI agent runtime.
Mainly Rust. The stack also includes Rust, Electron, Node.js.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.