Model or dataset
Stack-Cairn/LiveAgent avatar
Stack-Cairn/LiveAgent

Stack-Cairn/LiveAgent: A Local-First AI Agent Desktop with a Remote Gateway

A fully functional AI Agent desktop client that supports Webui access and can be creatively customized and expanded!

2,201 stars250 forksTypeScriptMIT

At a glance

What is it?
LiveAgent is a Tauri 2 desktop client that gives an AI agent real file-system, Bash and scheduled-task access, plus a Go gateway so you can drive it from a browser. It is a young project, and the README is thinner than its feature list suggests.
Who is it for?
Adopt LiveAgent if you want a desktop agent that can actually touch the local file system and you are comfortable building or checking a release from a project whose last push was 2026-09-10. Do not adopt it if you need a hosted, multi-tenant support desk, or if you cannot read the source to confirm what a tool call will do.
Can I use it commercially?
Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 1 day ago.
What is it written in?
Mainly TypeScript, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on September 30, 2026, and from our analysis. They are not legal advice.

Editorial analysis

What LiveAgent actually solves, and for whom

Most chat clients stop at text. LiveAgent's README frames the project as a local-first AI agent desktop client that "deeply integrates large language model reasoning with local system tools", so the agent can read and write files, run Bash, and manage scheduled tasks. The intended user is a developer or power user on macOS, Windows or Linux who wants an agent with hands, not a hosted assistant in a browser tab. The second audience is remote: the Go gateway plus WebUI let you control the same local agent from any browser, which matters if the machine doing the work is a workstation or home server you are not sitting in front of. The repository is a pnpm and Cargo workspace with crates for the GUI, shared UI, and the gateway, so the split between desktop and server is structural rather than bolted on.

How the desktop, tools and gateway fit together

The architecture is visible in the repository layout. crates/agent-gui holds the Tauri 2 shell with a Rust backend and a React 19 frontend; crates/agent-ui is the shared UI package; crates/agent-gateway is a Go 1.25 service with its own web/ directory for the WebUI. The README describes local tool execution as the core: Read, Write, Edit and Delete for files, Glob and Grep for pattern and regex search, Bash for non-interactive commands with cwd and timeout, and ManagedProcess for supervising long-running processes such as dev servers. Sub-agents run in parallel with worktree isolation and automatic merging, and TunnelManager exposes local services publicly. On the extension side, the Tauri layer bridges stdio and http MCP servers, and Skills packages load on demand. The gateway speaks WebSocket with Protobuf, and the README states that a bounded sequence window replays short disconnects while desktop-side persistence acts as the safety net. That is a coherent design: the desktop owns execution, the gateway owns reach.

Installing LiveAgent and running a first task

The README points to GitHub Releases for installers, which GitHub Actions builds, signs and publishes. On macOS you download the DMG matching your chip: LiveAgent-<version>-macOS-aarch64.dmg for Apple Silicon or LiveAgent-<version>-macOS-x64.dmg for Intel, open it, and drag the app into Applications. The README notes the macOS installer is signed and notarized by Apple, so no manual security override is needed on first launch. Windows requires x64 and the WebView2 runtime, which the README says ships with Windows 11. Linux needs x86_64 and WebKitGTK 4.1, with Ubuntu 22.04+ or Debian 12+ given as examples. If you prefer to build from source, the workspace uses pnpm 10.32.1 and exposes a dev stack script:

bash
pnpm install
pnpm dev:stack

The dev stack script starts the local services; pnpm dev:stack:status reports what is running and pnpm dev:stack:logs tails output. For the gateway specifically, the Makefile defaults are the concrete reference:

bash
make dev-gateway
make dev-webui

Those targets use DEV_GATEWAY_HTTP_ADDR ?= :50052 and DEV_GATEWAY_TOKEN ?= dev-token, so a local gateway listens on port 50052 with a development token. The Dockerfile builds the WebUI with Node 22.17.1, compiles the gateway with Go 1.25, and runs it as an unprivileged user with home directory /var/lib/liveagent. Once a model is configured through the app's multi-model routing, a reasonable first task is to point the agent at a scratch directory and ask it to grep for a symbol, then edit a file, so you can watch the tool calls before trusting it with anything real.

Where LiveAgent is the wrong tool

The name is the first hazard. LiveAgent here is a desktop agent client, not a customer-support live chat product, and search traffic for "liveagent vs freshdesk" or "is liveagent a crm" is about a different, unrelated product. If you need a support desk with ticketing and agent queues, this repository is not that. The second limitation is operational: an agent with Write, Edit, Delete and Bash is an agent that can damage your working tree, and the README does not document a sandbox, an approval gate, or a rollback path for tool calls. Sub-agent worktree isolation and automatic merging is a mitigation for parallel work, but the merge step is exactly where an unattended agent can produce surprises. Third, the remote gateway is a real attack surface: the README describes WebSocket plus Protobuf and disconnect recovery, but does not document authentication, TLS termination, or token rotation, so exposing the gateway beyond localhost is something you must verify in the source first. Finally, the project is young. The last push was on 2026-09-10 and the newest release is v1.3.2 from the same day, so the API surface and file layout can still move under you.

How it differs from Claude Desktop and Codex app

The closest points of comparison are the desktop clients whose protocols LiveAgent supports: Claude Desktop and the Codex app. Those are vendor clients tied to one model family's account and terms. LiveAgent's README instead advertises multi-model routing across Claude (Anthropic), Codex (OpenAI) and Gemini protocols, with custom Base URL support for third-party compatible services, and the sponsor section names relay providers for exactly that use. The practical difference is where the agent runs and who owns the loop: LiveAgent keeps execution on your machine and adds a gateway so a browser elsewhere can drive it, while a vendor client keeps you inside that vendor's surface. The trade-off is maintenance. A vendor client is patched for you; LiveAgent is a workspace you may end up building, with Tauri, Rust, Go, pnpm and Cargo all in the dependency graph.

Licence, upgrades and what maintenance costs you

The repository is MIT licensed, which permits commercial use, modification and redistribution provided the copyright notice and permission notice are included; that is a summary of the licence text, not legal advice, and you should read LICENSE in the repository for the binding terms. The upgrade cost is the part the README does not address. There is no documented migration path between v1.3.0, v1.3.1 and v1.3.2, and no changelog is shown, so the release notes on each tag are the only record of what changed. Because the workspace pins exact versions through pnpm overrides and a Cargo.lock, a source build is reproducible, but the desktop app, the gateway and the WebUI are separate artifacts that can drift apart if you upgrade one and not the others. If you run the gateway as a container, the Dockerfile gives you a fixed base and a non-root user, which keeps the deployment boring. If you run the desktop app, you are relying on the signed release pipeline for updates.

Editorial conclusion

Adopt LiveAgent if you want a desktop agent that can actually touch the local file system and you are comfortable building or checking a release from a project whose last push was 2026-09-10. Do not adopt it if you need a hosted, multi-tenant support desk, or if you cannot read the source to confirm what a tool call will do. Before installing, verify the release asset matches your platform and chip, and confirm the gateway's default token handling in the source, because the README does not document it.

Frequently asked questions

What is LiveAgent?

It is a local-first AI agent desktop client built with Tauri 2, React 19, Rust and Go, according to the README. It runs an agent with local file-system and Bash tool access, supports MCP servers and Skills packages, and can be reached remotely through a Go gateway with a WebUI.

Is LiveAgent legit?

The repository is public, MIT licensed, and publishes signed and notarized macOS installers through GitHub Releases, with the last push on 2026-09-10. Whether that is enough for your use is a judgement about a young project, not something the README can settle.

Is LiveAgent a CRM?

No. Stack-Cairn/LiveAgent is an AI agent desktop client with local tool execution and a remote gateway, not a customer relationship or support product. Searches that pair LiveAgent with CRM or helpdesk comparisons refer to a different, unrelated product of the same name.

Can live agents be AI?

In this project the agent loop is driven by large language models, with the README describing multi-model routing across Claude, Codex and Gemini protocols. The desktop client is the thing that runs the loop and the local tools, so the AI and the local execution are the same application.

Is LiveAgent down?

The desktop application runs locally, so there is no hosted service for this repository to take down. The only network dependency named in the README is the model endpoint you configure, including custom Base URLs for third-party compatible services.

Official sources

  1. Issues
  2. License: MIT
  3. README
  4. Releases
  5. Stack-Cairn/LiveAgent on GitHub
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/stack-cairn-liveagent.svg)](https://hysenlabs.com/projects/stack-cairn-liveagent)