AI Commander 0.1.0 is here, bringing multiple AI coding agent backends, a Kanban task board, Git worktree isolation, and permission controls into one retro-style terminal interface.
Think Norton Commander aesthetics meets modern agentic coding.
Built in Rust with ratatui, crossterm, and Tokio, AI Commander is designed for developers who want to manage AI coding workflows without constantly switching between terminals and agent interfaces.
WHAT'S NEW IN AI COMMANDER 0.1.0?
• Multi-agent backend support: connect OpenCode, Hermes Agent, DeepSeek Harness (DSH), and agents using the Agent Client Protocol (ACP).
• Terminal chat interface: multiple sessions, streaming responses, attachments, model selection, transcript search, and keyboard navigation.
• Live status monitoring: track the active harness, model, context usage, token consumption, estimated costs, and session time.
• Kanban task management: organize work through Backlog, Ready, In Progress, Review, and Done, with states for blocked tasks and human intervention.
• Git worktree isolation: execute tasks in separate worktrees, keeping parallel development work away from your main checkout.
• Task dependencies: use DAG relationships to control execution order and preserve artifacts between tasks.
• Controlled merging: review and approve completed work before a fast-forward merge, without modifying your active checkout.
• Permission policy engine: allow, ask, or deny operations based on project trust, normalized paths, and explicit approval rules.
• Human-in-the-loop controls: sensitive actions can pause execution until a human approves them.
• Secure IPC: token-based authentication separates permissions for human operators and AI agents.
• Context compaction: monitor context usage across supported backends and visualize compaction as a retro DOS-style disk defragmentation process called DEFRAG.
• Unified task execution: a shared TaskRunner powers both the interactive TUI and background daemon.
• Responsive event handling: Tokio's select-based event loop handles updates without polling delays.
• Cross-platform CI: automated checks on Linux, Windows, and macOS.
• 156 automated tests: unit, integration, property-based, mocked HTTP, and ACP conformance testing.
WHY IS THIS INTERESTING?
The challenge with AI coding agents isn't just generating code.
It's coordinating multiple tasks, tracking dependencies, controlling permissions, keeping contexts manageable, and reviewing what agents actually produce.
AI Commander brings these concerns into one terminal-based workspace.
And the retro interface is a nice bonus.
But version 0.1.0 is still an early release.
Kanban teams aren't implemented yet. Some backend context metrics are estimates rather than exact measurements. Prebuilt executables aren't available in the initial release, so you'll need to compile the project from source.
The current release supports four harness backends, not every AI coding agent automatically.
AI Commander is free and open source under the MIT license.
GitHub:
https://github.com/mb-mal/ai-commander
Version 0.1.0 release:
https://github.com/mb-mal/ai-commander/releases/tag/v0.1.0
Created by Mikhail Malyshev.
Would you rather manage coding agents through a modern graphical interface or a powerful retro-style terminal?
#AICommander #AICoding #OpenSourceAI
YouTube / ABVX
