Support

AgentOrcha is an iOS companion app for the Agent Orcha framework. It lets you chat with AI agents and local language models, connect to other Agent Orcha instances over a peer-to-peer network, and manage your P2P settings — all from your iPhone or iPad.

AgentOrcha requires iOS 17.0 or later.

An internet connection is only needed to download AI models from Hugging Face. Once downloaded, local LLM inference runs entirely on-device. P2P connections work over any network where peers can reach each other (local Wi-Fi or internet).

Go to the Network tab to see available peers. The app uses the Hyperswarm protocol to discover other Agent Orcha instances sharing the same network key. You can configure your peer name and network key in Settings. Use the default key to find public peers, or set a custom key for a private network.

Try the following:

• Make sure at least one other Agent Orcha instance is running and connected to the same network key.
• Check that your device has an active network connection.
• If using a custom network key, verify that all peers are using the exact same key.
• On corporate or restricted networks, the Hyperswarm protocol may be blocked. Try switching to a different network.

Go to the Local LLM tab, browse the available models, and tap Download on the one you want. Once the download completes, select the model and start chatting. Models run entirely on your device — no data is sent externally.

Local AI inference depends on your device's hardware. The app automatically selects the best model for your device based on available RAM. Tips:

• Close other apps to free up memory.
• Ensure your device is not in Low Power Mode, which may throttle processing.
• Newer devices with more RAM and a faster Neural Engine will perform significantly better.
• The app requires an A14 chip or newer (iPhone 12+).

We do not operate servers, collect user data, or track usage. However, there are important distinctions depending on how you use the app:

• Local LLM: Fully private. All processing happens on your device and no data leaves it.
• P2P chat with remote agents or models: Your messages are encrypted in transit using the Noise protocol, protecting them from network eavesdropping. However, the host peer that processes your request can see your messages — this is necessary for the host to forward them to the LLM for inference. This is the same trade-off as using any remote AI service. Only connect to peers you trust.

See our Privacy Policy for full details.

Go to the Local LLM tab, find the model you want to remove, and use the delete option. You can also remove all downloaded models by deleting and reinstalling the app.

The quickest way is via npx:

npx agent-orcha@latest

You can also run it with Docker or download the standalone binary from the GitHub Releases page. See the Quick Start Guide for full instructions.

Agent Orcha is model-agnostic and supports OpenAI, Anthropic, Google Gemini, Ollama, LM Studio, local models via Omni, and remote models via the P2P network. Configure providers in models.yaml.

Open an issue on our GitHub Issues page. Please include:

• A clear description of the problem
• Steps to reproduce
• Your platform and version (iOS version, device model, or framework version)
• Any relevant error messages or screenshots

Yes. Agent Orcha is 100% open source and released under the MIT License. The iOS app and framework are free to use.