Kuddev/pebrel

AI-native, GPU-accelerated terminal emulator for Windows with SSH, persistent sessions, split panes, and first-class AI CLI workflows.

What it solves

Pebrel is a GPU-accelerated terminal emulator and workspace designed specifically to streamline workflows for users of AI command-line interface (CLI) tools like Claude Code and Codex. It solves the problem of fragmented workflows by integrating local shells, remote SSH hosts, file browsing, and AI agent activity tracking into a single native desktop application.

How it works

Built with Rust and the GPUI framework, Pebrel provides a high-performance interface that manages multiple terminal sessions in tabs and split panes. It integrates with AI CLIs by recognizing specific tools and using hook events to provide visual signals for agent activity and progress. It also includes a native document reader that renders Markdown and mathematical formulas without requiring a WebView, allowing AI-generated answers to be viewed in a dedicated reader tab.

Who it’s for

Developers and power users who rely on AI-powered CLI agents and remote server management via SSH/SFTP, and who want a consolidated workspace for their terminal sessions.

Highlights

  • GPU-accelerated interface with customizable themes and opacity.
  • Native support for Claude Code and Codex with dedicated icons and activity tracking.
  • Integrated SSH and SFTP client with support for jump-hosts and private keys.
  • Native rendering of Markdown and mathematical formulas for AI output.
  • Lua-based configuration with live reload and validation.
  • Support for Windows, macOS, and Linux.

Related

  • Project
    doobidoo/mcp-memory-service

    A persistent, shared memory backend for AI agents that provides a knowledge graph and semantic search to maintain context across sessions and different AI tools.

  • Project
    alvinunreal/tmuxai

    An intelligent terminal assistant for tmux that observes pane content and executes commands to act as a context-aware pair programmer.

  • Project
    CannyLab/tsne-cuda

    A GPU-accelerated implementation of the FIt-SNE algorithm that provides massive speedups for visualizing high-dimensional data compared to CPU-based methods.

  • Project
    meta-pytorch/opacus

    A PyTorch library for training machine learning models with differential privacy, allowing users to protect sensitive training data while tracking privacy loss.