egoist/lorca

Imagine Telegram but single person, with agents, and end-to-end encrypted. Alternative to Grok bot, Muse, Dots

What it solves

Lorca offers a private, self-hosted alternative to cloud bot services like Grok. It lets you run bots on your own machines, keep your conversations end-to-end encrypted, and have multiple bots collaborate and hand work to each other—all without relying on a central service for your data.

How it works

The system is split into parts: a Rust CLI that acts as a local websocket service and runs the agent loop, a Go desktop app that provides a native UI for that CLI, and phone apps (built with Expo) that pair as devices. On first use, each device generates a local key pair. Devices pair to each other via QR code or a pairing string. All traffic to the network relay is end-to-end encrypted using X25519/Ed25519 key exchange and XChaCha20-Poly1305 for data. Desktop devices are "Runners"—they can host bots and use encrypted provider credentials. Phones and tablets are "Devices" that can read and write chats, attach files, and start bot turns, but the bots actually run on Runners.

Who it's for

Individuals or teams who want a privacy-first, self-hosted bot platform with control over where bots run and how data flows. It's also for people who want to orchestrate multiple AI bots in group chats, similar to Grok Bot, but with their own infrastructure.

Highlights

  • Cross-platform support: desktop apps for macOS, Windows, and Linux, plus phone apps for iOS and Android.
  • End-to-end encryption with modern crypto (X25519/Ed25519, XChaCha20-Poly1305).
  • Bots can hand work to each other and orchestrate in group chats.
  • Pair your phone to control bots running on your desktop machines.
  • Supports multiple AI providers: DeepSeek, Anthropic, OpenCode Zen/Go, ChatGPT, and Grok (SuperGrok or X Premium+).
  • Desktop app uses native UI for each system, with materials, density, keyboard, and motion that feel at home.

Related

  • Project
    Zero-Mem/Zero-mem

    Zero-Mem is a memory framework for LLM agents that performs all memory operations (construction, organization, retrieval, calibration) without any LLM calls or tokens, using only an encoder and NER for computation. It builds an entity–context graph and temporal hierarchy over original interaction traces, retrieves evidence via query-conditioned routing, and uses a reader LLM only for final answer generation. Reported results show GPT-4o-mini achieving 59.15 F1 on LoCoMo and 72.07 F1 on HotpotQA 56K, with 57.6% lower memory-operation time than the LightMem baseline.

  • Project
    WellApp-ai/Well

    A Chrome extension that uses AI to automate the retrieval of supplier invoices from web portals and messaging apps, piping data directly into accounting tools.

  • Project
    mediar-ai/fazm

    Fazm is an open-source AI computer agent that lets you control your browser, write code, manage documents, and operate Google Apps via voice, learning your workflows along the way. It's free to start, fully local, and runs as a macOS app.

  • Project
    yjh051108/dsh-router-standard

    A research‑level router preset for DeepSeek Harness that automatically classifies the first user request and injects the appropriate persona and tool set (spec/react/weak) for the DeepSeek V4 model. It runs as a zero‑dependency Cord‑is plugin, integrates with the dsh‑super‑injector for hot‑reloading, and provides CLI diagnostics. Currently in development (v1.19.1, Aug 2026).