Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
Overview & Reference

Nekoni Review: A Self-Hosted AI Agent You Reach From Your iPhone

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

Nekoni (iPhone app 2.0, repository updated on 24 October 2026) is a self-hosted AI agent that runs on a computer you own and that you chat with from an iPhone or a browser. It is free and MIT-licensed, installs with one terminal command on macOS or Linux, and every statement here comes from its website, repository and App Store listing.

Key Takeaways

  • Maintenance: the repository received commits through 24 October 2026, but its only tagged release is v0.1.0 from 27 March 2026 and the iPhone app 2.0 dates from 6 April 2026, so no stale-project warning applies.
  • Install: one command on macOS or Linux (Windows through WSL2) that sets up Docker, Ollama and a model of roughly 4 GB.
  • Clients: an iPhone app on the App Store and a browser app; Android is listed as coming soon.
  • Cost: free, with the code under the MIT license.
  • Maker: not named on the website; the App Store seller and the commit author are one individual.

📍 In One Sentence

Nekoni is an open-source agent that you install on your own computer, with an iPhone app and a web app as remote controls.

💬 In Plain Terms

You run one install command at home, scan a QR code with your phone, and then chat with a local model, ask questions about your documents and schedule recurring prompts from anywhere.

📌Note: This review is based on the website, repository and App Store listing, checked on 24 October 2026. PromptQuorum has not installed or benchmarked Nekoni, so speed, answer quality and connection reliability are not assessed.

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

Nekoni is an MIT-licensed, self-hosted AI agent: you run it on your own Mac or Linux computer with Ollama, then chat with it from an iPhone app or a browser over a direct WebRTC link. This review is based on the project's website, GitHub repository and App Store listing, checked on 24 October 2026; PromptQuorum has not tested it hands-on.

Nekoni Review: A Self-Hosted AI Agent You Reach From Your iPhone

What Is Nekoni?

Nekoni is a self-hosted AI agent made of a Python agent, a web dashboard and a signalling server, with a React Native iPhone app and a browser app as clients. The agent uses Ollama for inference, ChromaDB for search over your documents and SQLite for its data, and the repository describes itself as a monorepo of five apps.

The website does not name a maker. The App Store lists the seller as Denys Malykhin, the same name appears as the author of the repository's commits, and his GitHub profile points to nekoni.dev. The same GitHub organization also publishes a separate project called murmur, a Bluetooth shell for a Raspberry Pi, which is a different product from the Murmur AI music app reviewed on PromptQuorum.

Get It

Nekoni has two parts: an agent you install on a Mac or Linux computer with a one-line terminal command, and clients (an iPhone app and a browser app) that connect to it. Android is listed as coming soon.

Agent (macOS, Linux)

How to get it:
Run curl -sSL https://raw.githubusercontent.com/nekonihq/nekoni/refs/heads/main/install.sh | bash in a terminal

iPhone

How to get it:
Nekoni on the App Store, free

Browser

How to get it:
app.nekoni.dev, a web app that pairs with the same agent

Android

How to get it:
Listed as "coming soon" on the website; no store listing found

Website

How to get it:
nekoni.dev

Privacy policy

How to get it:
PRIVACY.md in the repository, last updated on 21 March 2026; the website has no separate privacy page

Source code

How to get it:
github.com/nekonihq/nekoni, MIT license

A terminal and administrator rights are required: the installer adds Docker, uv and Ollama when they are missing. The iPhone app is a normal App Store install but does nothing useful without the agent running at home. Version reviewed: iPhone app 2.0 (released on 6 April 2026, 47.9 MB, iOS 15.1 or later); the repository has one tagged release, v0.1.0, with no downloadable binaries.

Requirements

The agent needs macOS or Linux (Windows through WSL2), Docker and about 4 GB of free disk space for the default model.

Operating system

What the sources say:
macOS or Linux for the agent; Windows through WSL2 per the README. The installer's Windows hint points to an install.ps1 script that returned a 404 error at nekoni.dev on 24 October 2026

Software

What the sources say:
Docker (installed by the script through Homebrew on macOS or get.docker.com on Linux), uv and Ollama; on macOS the script stops if Homebrew is missing

Model

What the sources say:
llama3.2 by default, about 4 GB of disk; changeable through the OLLAMA_MODEL setting

iPhone

What the sources say:
iOS 15.1 or later per the App Store; the camera scans the pairing QR code

Network

What the sources say:
Pairing happens once on the local network; afterwards phone and agent find each other through the project's public signalling server

The documentation gives no RAM, GPU or CPU requirement, so how large a model your computer can run depends on the Ollama model you choose and is not stated.

How to Get Started

The steps follow the project's README and setup guide; PromptQuorum has not run them.

  1. 1
    Run the installer
    Why it matters: In a terminal on your Mac or Linux computer, run the install command from the Get It table; it pulls the default model and starts the agent and dashboard.
  2. 2
    Open the dashboard
    Why it matters: Open the dashboard on port 8080 of that computer and change the default login; the documentation lists admin and nekoni as the defaults and says to change the password.
  3. 3
    Pair your phone
    Why it matters: Open the Nekoni app, scan the QR code on the dashboard's Pair page and approve the request in the dashboard.
  4. 4
    Chat, upload, automate
    Why it matters: Chat with the model, upload documents to the knowledge base, and create skills that you run on demand or on a schedule.

Features Confirmed by the Sources

Every item below comes from the website, README or App Store description; none has been independently tested.

  • Chat. Talks to models served by Ollama and keeps a conversation history you can continue; the App Store text names Llama, Mistral and Gemma as examples.
  • Knowledge base (RAG). Upload PDF, Word, Excel, PowerPoint, HTML and image files; Docling extracts the text, including OCR, and ChromaDB indexes it locally for questions in chat.
  • Skills and cron. Skills are named prompt templates that run on demand or on a schedule, for example a daily summary.
  • Extensible tools. A short Python file dropped into the tools directory becomes a new capability for the agent.
  • Dashboard and traces. A web dashboard handles pairing, the knowledge base and skills, and shows a live trace stream of what the agent does.

Data Flow and Privacy

The privacy policy, last updated on 21 March 2026, says the app and the developer collect no messages, name, email, location, analytics or advertising data. Chat history is stored on the iPhone, the Ed25519 identity key sits in the iOS Keychain, and documents and logs stay on the machine that runs the agent.

The project describes one public component, a signalling server that it operates. According to the policy it exchanges only the WebRTC set-up data (SDP and ICE) and does not log or store data beyond the active session; after connecting, messages travel over an encrypted channel between phone and agent. The security documentation adds that every signalling message is signed and that a four-step mutual authentication runs before the agent accepts a message. The camera is used only to scan the pairing QR code.

📌Note: These are the project's own declarations, not audit results, and PromptQuorum has not inspected the network traffic. The policy speaks of the mobile app and does not mention the hosted browser app.

Who Should Use It

  • Self-hosters and homelab owners who are comfortable with Docker and want a private assistant reachable from their iPhone.
  • Ollama users who want a phone front end plus document search and scheduled prompts on top of the models they already run.
  • Developers who want to extend an MIT-licensed agent with their own Python tools.

What We Could Not Verify

  • Speed and quality. PromptQuorum did not run the agent, so answer quality, response time and the load on your computer are not assessed; the documentation names no minimum hardware.
  • Connection reliability. The project says phone and agent connect directly; whether this works from other networks was not tested.
  • Security. The authentication design is documented, but no independent audit is listed in the repository or on the website, and PromptQuorum has not reviewed the code.
  • Project maturity. The repository was created in March 2026, shows a single tagged release and lists one contributor, so long-term maintenance is unknown.
  • Ratings. The App Store listing showed no user ratings on 24 October 2026.

Competitors and Alternatives

LocalGhost

Platforms:
Linux server, Android app
Price / license:
Free / MIT
Key difference:
Personal AI server that keeps photos, notes and voice notes, with an Android app as the only client
Articles about LocalGhost (1)

Odysseus

Platforms:
Windows, macOS, Linux
Price / license:
Free / AGPL-3.0
Key difference:
Self-hosted AI workspace for chat, agents, research and documents, used from a desktop app or the browser
Articles about Odysseus (2)

Also mentioned in:

Khoj

Platforms:
Windows, macOS, Linux, web
Price / license:
Free / AGPL 3.0
Key difference:
Second brain that searches your notes and documents and syncs with tools such as Obsidian
Articles about Khoj (4)

Also mentioned in:

+4 more not shown

Confirm each app's current price, license and platforms on its own listing.

Compare in directory

Frequently Asked Questions

Does Nekoni work on Windows or Android?

The README supports Windows for the agent only through WSL2, and the website lists the Android app as coming soon without a date. The browser app is the alternative to the iPhone app.

Which model does Nekoni use?

The default is llama3.2, pulled through Ollama. Changing the OLLAMA_MODEL setting and running make pull switches to another Ollama model; PromptQuorum did not test other models.

Does it need an internet connection?

Inference and storage are local, but phone and agent find each other through the project's public signalling server, so both need internet access when you connect from outside your home.

How do I stop or restart it?

The README gives cd ~/.nekoni && make down to stop the agent and cd ~/.nekoni && make up to start it again.

Verdict

Nekoni is a lean, MIT-licensed way to put a local Ollama agent, document search and scheduled prompts behind an iPhone app, with a one-line install and no subscription. Against that, it is a young project that appears to come from one developer, its only tagged release dates from March 2026, the mobile client exists only for iPhone, and its privacy and security statements rest on the project's own documents and were not tested here. It suits self-hosters who already run Docker and Ollama and want remote access; anyone who needs an audited or Android-ready product should wait or check the repository first.

Sources

  • Nekoni on the App Store — description, version, seller, requirements and price, checked on 24 October 2026.
  • Nekoni website — feature list, install command and platform notes, checked on 24 October 2026.
  • Nekoni on GitHub — README, install script, setup, configuration and security documents, license and commit history, checked on 24 October 2026.
  • Nekoni privacy policy — data collection, signalling server and camera use, last updated on 21 March 2026.

← Back to Power Local LLM

MCP