Logician

A local-first coding agent with a streaming terminal UI. SSH-ready, thinking-visible, and built for real code editing workflows.

Logician turns natural-language instructions into inspectable code changes with streamed progress, session persistence, and skill-based extensibility. It works with local or hosted OpenAI-compatible model endpoints.

Get StartedRead the GuidesView on GitHub

Features

• Terminal-native — works over SSH, in tmux, in any VT100-compatible terminal
• Streaming responses — see reasoning, tool progress, and results as they happen
• Safe edits — strict text matching, CRLF/BOM preservation, path normalization
• Skills & plugins — SKILL.md-driven capabilities, Claude-style lifecycle hooks
• Subagents — delegate bounded tasks to child agents
• Optional structured reasoning — SSR, Tree of Thoughts, Reflexion, and more
• Session management — persistence, bookmarks, branching, rewind checkpoints, compaction
• Cross-session memory — persistent observations, lessons, and action tracking
• MCP support — stdio and streamable HTTP MCP servers
• Permission modes — acceptAll, acceptEdits, ask, plan

Quick Start

git clone https://github.com/lseman/logician.git
cd logician
bun install
bun start

The TUI connects to an OpenAI-compatible backend at http://127.0.0.1:8080 by default. Configure it in ~/.logician/settings.json, a trusted project's .logician.json, or via LOGICIAN_LLM_URL.

Architecture at a glance

What can it do?

Capability	Description
Code editing	Safe, verified edits with exact text matching
Reasoning	Tree of Thoughts, SSR, Reflexion, Auto-CoT
Session persistence	Full history, branching, rewind, compaction
Skills	SKILL.md-driven capabilities loaded automatically
Subagents	Delegate tasks to child agents
MCP integration	Tool extension via MCP servers
Headless mode	JSONL streaming for CI/CD pipelines