Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Local AI Agents & Tool Use

Local Operator Review: Free Desktop AI Agent for Your Computer

·7 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

Local Operator (desktop app version 0.42.2, released on 24 October 2026) is a free, MIT-licensed desktop app whose AI agents write and run code on your own computer. The agent and your files stay local, but the model can be hosted, in which case it receives your conversation. Every statement here comes from the maker's site, its GitHub repositories and its privacy policy.

Key Takeaways

  • Maintenance: the desktop app (0.42.2) and the agent runtime (0.88.3) both shipped releases on 24 October 2026, so no stale-project warning applies.
  • Models: local servers such as Ollama, LM Studio, vLLM and llama.cpp, or hosted providers through your own keys or Radient Pass.
  • Safety: reads run automatically, while file writes and shell commands ask first and show the exact command.
  • Cost: the app is free and open source; per the maker, the only thing you might pay for is AI model usage.
  • Maker: Radient Inc., with both repositories published under the MIT license.

📍 In One Sentence

Local Operator is a free, open-source desktop app in which AI agents write and run code on your computer, using a local Ollama model, your own API keys or the maker's hosted Radient Pass.

💬 In Plain Terms

You describe a job such as sorting a folder of invoices, the app shows a plan, a second agent checks it, you approve, and the code runs on your machine.

📌Note: This review is based on the maker's website, its GitHub repositories and its privacy policy, checked on 24 October 2026. PromptQuorum has not installed or run the app, so task quality and speed are not assessed.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

Local Operator is a free, MIT-licensed desktop app from Radient Inc. in which AI agents plan a task, have the plan checked by a separate agent and then write and run code on your own Mac, Windows or Linux computer. The model behind it can be a local Ollama model, your own API keys or the maker's hosted Radient Pass. This review is based on the maker's website, its two GitHub repositories and its privacy policy, checked on 24 October 2026; PromptQuorum has not tested the app hands-on.

Local Operator Review: Free Desktop AI Agent for Your Computer

What Is Local Operator?

Local Operator is an open-source agent app from Radient Inc. that turns a plain-language request into a plan and then runs real code on your computer to carry it out. It has two parts: the desktop app in the repository local-operator-ui and the Python agent runtime in local-operator, which also runs as a terminal app called lop.

The site presents it as a team of agents: you can save agents, group them into teams with a manager, schedule recurring work and borrow shared agents from the Agent Hub, which is powered by Radient. The core repository was created on 25 January 2025.

Get It

Local Operator is a free download for macOS, Windows and Linux (x86_64); the installers are hosted on GitHub Releases.

Option
How to get it
macOSDesktop app on GitHub Releases: .dmg for Apple silicon and for Intel
WindowsSame release page: .exe installer for x64, for Arm64 or a combined one
Linux (x86_64)Same release page: .AppImage, .deb and .rpm; there is no Arm64 Linux build
Terminalnpx local-operator-ui (needs Node.js 22.13.1 or newer)
Websitelocal-operator.com
Privacy policylocal-operator.com/privacy-policy
Source codeMIT license: local-operator (agent runtime) and local-operator-ui (desktop app)

Version reviewed: desktop app 0.42.2, released on 24 October 2026 per its GitHub release, on top of agent runtime 0.88.3. The website's download page still shows version 0.12.8 from 11 February 2026 and macOS files named 0.19.7, so it is out of date; GitHub Releases is newer and is used here. Both repositories published several releases on 24 October 2026.

Requirements

The download page lists macOS 11 or newer, Windows 10 or newer and a recent x86_64 Linux, plus a model to talk to. It states no RAM or VRAM minimum.

macOS

What the sources say:
macOS 11 (Big Sur) or newer, Intel or Apple silicon

Windows

What the sources say:
Windows 10 or newer, x64 or Arm64

Linux

What the sources say:
Recent x86_64 distribution, for example Ubuntu 18.04+, Debian 10+ or Fedora 32+; the .deb and .rpm install Python 3.12 packages

Disk

What the sources say:
A few hundred megabytes for the installer plus room for the backend, which the app sets up on first run

Model

What the sources say:
A local Ollama model (free, works offline), your own API keys or a Radient Pass balance

The page predates the current releases, so check the system requirements again when you download. How much memory a local model needs depends on the model you choose, not on the app.

How to Get Started

The steps follow the maker's download page and first-run description; PromptQuorum has not run them.

  1. 1
    Install the app
    Why it matters: Run the installer for your system. On macOS the first launch is blocked until you choose Open Anyway under System Settings, Privacy & Security; Windows may show a SmartScreen warning with More info, Run anyway.
  2. 2
    Let it set up
    Why it matters: On first run the app installs its own backend, which takes a minute or two, and opens straight into a chat.
  3. 3
    Connect a model
    Why it matters: Choose Ollama, your own API keys or Radient Pass; you can switch per agent or per conversation.
  4. 4
    Ask for something small
    Why it matters: The agent shows its plan, a separate agent checks it, and you approve before anything runs.

Features Confirmed by the Sources

Every item below comes from the maker's website or the project README; none has been independently tested.

  • Plan, check, approve. The agent shows a plan, a separate security-checking agent reviews each proposed action, and you sign off. Per the README, reads run automatically while writes and shell commands show the exact command and ask first.
  • Writes its own tools. When a task needs a capability that does not exist, the agent writes code to provide it, and it shows each command and its output as a receipt.
  • Teams and subagents. Agents can be grouped into teams with a manager, and work can be split across subagents that run in parallel, each showing its elapsed time, context use and cost.
  • Schedules and memory. Recurring tasks run in the background on a schedule, and an agent keeps what you teach it for later requests.
  • Desktop extras. A Chromium browser pane with an approval prompt before the agent acts on a page, and a file canvas that opens markdown, code, PDFs, spreadsheets, images, audio and video.
  • Connections. MCP servers are supported, and devices you own can be paired into a mesh so sessions are reachable from each of them.

Data Flow and Privacy

The maker says your files never have to leave your computer, while the text of your conversation goes to whichever model you pick. With a local Ollama model nothing leaves; with Radient Pass or your own API keys the conversation goes to that provider under its terms. No account is required, and one is needed only for the optional cloud platform.

Web search is on by default according to the README, and can be switched off with lop search off; fetching web pages has its own switch. By default a command the agent runs can see your environment variables. The privacy policy, effective 4 April 2025, says Radient may collect device, usage and log data and, for cloud accounts, name and email.

📌Note: These are the maker's declarations, not audit results. PromptQuorum has not inspected the app's network traffic or confirmed which data each setting stops.

Who Should Use It

  • Desktop users with file chores such as sorting, renaming or summarising documents, who want an agent that asks before it writes.
  • Ollama users who want a graphical agent app that can run on a local model.
  • Tinkerers who value an MIT-licensed project with open repositories and frequent releases.

What We Could Not Verify

  • Task quality and speed. PromptQuorum did not run the app, so the maker's examples, such as the invoice job, are not reproduced.
  • Local-model results. The sources give no minimum model or hardware for useful agent work, and local models were not tried.
  • The security check. How reliably the separate checking agent catches a harmful command is the maker's description, not a tested property.
  • Code signing. The download page warns of macOS and Windows prompts; whether the installers are signed was not checked.
  • Version drift. The website still shows version 0.12.8 and the repositories release several times a day, so features can change between your download and this review.

Competitors and Alternatives

Open Interpreter

Platforms:
Windows, macOS, Linux
Price / license:
Free / Apache 2.0
Key difference:
Lets a local model control your computer and run code; check its own review for install path and license terms
Articles about Open Interpreter (3)

Also mentioned in:

Agent Zero

Platforms:
Windows, macOS, Linux
Price / license:
Free / MIT
Key difference:
Self-hosted agent that runs in Docker with a web interface rather than a native installer
Articles about Agent Zero (2)

Also mentioned in:

Nanobot

Platforms:
Windows, macOS, Linux
Price / license:
Free / MIT
Key difference:
Self-hosted personal agent with persistent memory, MCP tools and chat-app integrations, used from a web or command-line interface

These are the PromptQuorum reviews in the same segment; confirm each app's current price, license and platforms on its own site.

Compare in directory

Frequently Asked Questions

Can Local Operator run fully offline?

Per the maker, yes with a local Ollama model, which it calls the most private option and says works offline. Web search is on by default, so switch it off if nothing should leave the machine.

Why does the download page show an older version?

The page still lists version 0.12.8 from 11 February 2026 and macOS files named 0.19.7, while GitHub Releases shows desktop app 0.42.2 from 24 October 2026. Take installers from the GitHub release page.

Will the agent change files without asking?

By default it asks first. The README says reads run automatically, writes and shell commands show the exact command, and turning approvals off is an explicit choice.

What does Radient Pass cost?

The sources give no price here: Radient Pass is described as an optional prepaid balance across hosted providers, billed by compute used. Check the maker's page before relying on it.

Verdict

Local Operator is an actively released, MIT-licensed way to run a code-writing agent on your own computer, with an approval step before anything is written or executed and the option to use a local Ollama model. Against that, it is untested here, a hosted model sees your conversation, the website lags the releases, and the project ships so often that this review will age quickly. It suits Mac, Windows and Linux users who want a graphical agent for their own files and will check the version, the model setting and the web-search setting before trusting it with anything sensitive.

Sources

← Back to Power Local LLM

MCP