computerex/z

Agentic harness optimized for z.ai GLM models

6

stars

210

commits

Python

primary language

Aug 28, 2026

updated

README

z

z terminal

An agentic coding harness for the terminal. Bring your own API keys — hot-swap models mid-session with /providers use. No subscriptions.

Quick start

pip install -e .
z --install   # one-time setup wizard
z

Config lives in ~/.z.json:

{ "api_url": "https://api.z.ai/api/paas/v4/", "api_key": "sk-...", "model": "glm-4.7" }

Highlights

  • 26 providers, one client — Z.AI, DeepSeek, Anthropic, OpenAI, Ollama, Bedrock, Copilot OAuth, OpenRouter, all via LiteLLM.
  • Constant streaming — every token, including reasoning, streams live. No spinners.
  • Telegram remotez --telegram <token> and drive the agent from your phone.
  • Context that survives compaction — half/quarter/lastTwo strategies, todo persistence, tool-result caching.
  • Claude Code semantics — CLAUDE.md, hooks, persistent memory, cron tasks, sub-agents, MCP.
  • 19 built-in tools — file ops, shell, search, web search, image analysis, background processes.

Commands

CommandWhat
/providers use <name>Switch models mid-session
/compactForce context compaction
/costToken usage and spend
/todoTodo panel
!cmdRun shell command
Ctrl+TToggle reasoning effort

Screenshots

z setup wizard

z provider selection

z model selection

Plugins: drop a .py in ~/.z/plugins/.

MIT

Contributors

computerex

210 commits

computerex/z

Agentic harness optimized for z.ai GLM models

6

stars

210

commits

Python

primary language

Aug 28, 2026

updated

README

z

z terminal

An agentic coding harness for the terminal. Bring your own API keys — hot-swap models mid-session with /providers use. No subscriptions.

Quick start

pip install -e .
z --install   # one-time setup wizard
z

Config lives in ~/.z.json:

{ "api_url": "https://api.z.ai/api/paas/v4/", "api_key": "sk-...", "model": "glm-4.7" }

Highlights

  • 26 providers, one client — Z.AI, DeepSeek, Anthropic, OpenAI, Ollama, Bedrock, Copilot OAuth, OpenRouter, all via LiteLLM.
  • Constant streaming — every token, including reasoning, streams live. No spinners.
  • Telegram remotez --telegram <token> and drive the agent from your phone.
  • Context that survives compaction — half/quarter/lastTwo strategies, todo persistence, tool-result caching.
  • Claude Code semantics — CLAUDE.md, hooks, persistent memory, cron tasks, sub-agents, MCP.
  • 19 built-in tools — file ops, shell, search, web search, image analysis, background processes.

Commands

CommandWhat
/providers use <name>Switch models mid-session
/compactForce context compaction
/costToken usage and spend
/todoTodo panel
!cmdRun shell command
Ctrl+TToggle reasoning effort

Screenshots

z setup wizard

z provider selection

z model selection

Plugins: drop a .py in ~/.z/plugins/.

MIT

Contributors

computerex

210 commits

Languages

Python

98.1%

Go

1.9%