GitHub v1.0.0
thwip-cli v1.0.0 is live on PyPI.

The open source AI coding agent multiplexer

Detect all installed AI coding agents on your machine, combine them in a unified terminal REPL, and hot-swap models mid-conversation without losing context, files, or execution state.

$ pip install thwip-cli
Anthropic Claude thwip ~/Developer/Projects/backend [session: auth-refactor]
claude | claude-sonnet-4 | [chat] [edit] [run] [term] [git]
1,482 tok | $0.0084

How thwip Works

A unified multiplexing pipeline that turns fragmented agents and models into one cohesive, resilient terminal development environment.

  • [*]
    1. Zero-Config Discovery

    On startup, thwip automatically scans your machine for installed CLI binaries (claude, agy, gemini, codex, aider, ollama), IDE installations (Cursor, Windsurf, VS Code), and existing credentials in ~/.claude.json, ~/.gemini/config.json, and environment variables.

  • [*]
    2. Unified Cross-Provider State

    Conversations are preserved in a standardized message format. When you switch between Claude, Gemini, OpenAI, DeepSeek, or Ollama, the complete history, user prompts, tool outputs, and reasoning steps migrate seamlessly.

  • [*]
    3. Shared Universal Tool Engine

    Every active agent gets direct, safe access to the same local capabilities: reading files, structured regex editing, creating files, running shell commands, executing sandboxed tests, and committing Git operations.

  • [*]
    4. Mid-Session Model Hot-Swapping

    Switch models at any point during your workflow using /switch <agent> <model> or the interactive Ctrl+S picker without starting a new session or losing modified workspace files.

  • [*]
    5. Automatic 429 Rate Limit Failover

    When an API returns an HTTP 429 rate limit or quota exhaustion error, thwip intercepts the failure, preserves the turn, and offers an immediate one-key fallback to a ready secondary provider.

  • [*]
    6. Offline & Local Ollama Support

    Full integration with local Ollama daemons. Code completely offline with DeepSeek-R1, Qwen, or Llama models with zero API keys and zero cost.

Architecture & Protocols

Engineered for high availability, zero context loss, and deterministic tool execution.

Fig 1.
Universal Multiplexing
Seamless state migration across provider SDKs.
Fig 2.
Unified Tool Engine
Deterministic file editing, test execution, and git operations.
Fig 3.
429 Failover Interceptor
Instant fallback switching when quotas exhaust.

Supported Providers & Models

Native adapter layer supporting the world's leading coding models.

Provider Default Models Auto-Discovery Location Context Limit Tools Supported
Anthropic claude-sonnet-4, claude-opus-4, claude-3-5-haiku ~/.claude.json, ANTHROPIC_API_KEY 200K tokens Chat, Edit, Run, Term, Git
Google Gemini gemini-2.5-pro, gemini-2.5-flash, gemini-2.0-flash ~/.gemini/config.json, GEMINI_API_KEY 1M+ tokens Chat, Edit, Run, Term, Git
OpenAI gpt-4.5-preview, gpt-4o, o3-mini, o1 ~/.config/openai/, OPENAI_API_KEY 128K-200K tokens Chat, Edit, Run, Term, Git
DeepSeek deepseek-chat, deepseek-reasoner (R1) DEEPSEEK_API_KEY, ~/.thwip/config.toml 64K tokens Chat, Edit, Run, Term, Git
Groq llama-3.3-70b-versatile, deepseek-r1-distill-llama-70b GROQ_API_KEY, ~/.thwip/config.toml 128K tokens Chat, Edit, Run, Term, Git
Ollama (Local) deepseek-r1:latest, qwen2.5-coder:32b, llama3.3 http://localhost:11434 (Auto-detected) Local GPU/VRAM Chat, Edit, Run, Term, Git
OpenRouter anthropic/claude-3.5-sonnet, meta-llama/llama-3.3-70b OPENROUTER_API_KEY, ~/.thwip/config.toml Provider-dependent Chat, Edit, Run, Term, Git

Interactive Navigation & Commands

Full reference of slash commands and keyboard shortcuts inside the thwip REPL.

Command / Shortcut Arguments Description Example
/switch <agent> [model] Hot-swaps active agent and model mid-conversation with zero lost context /switch google gemini-2.5-pro
/agents None Lists all discovered local agents, configured API keys, and active status /agents
/models [agent] Displays available models, context window limits, and pricing for a provider /models openai
/about or /guide None Opens the in-depth system guide and navigation manual inside the terminal /about
/limits None Inspects live session tokens, cumulative costs, and rate limit quotas /limits
/tools None Lists available universal tools (read, edit, write, run, bash, git) /tools
/history None Shows conversation turn summary and token breakdown across all models /history
/session save <name> Persists the current conversation history and file changes to disk /session save refactor-auth
/session load <name> Restores a saved session and resumes context from where you left off /session load refactor-auth
/clear None Clears current conversation memory and resets session tokens /clear
Ctrl + S Hotkey Interactive agent and model switch prompt popup Hotkey
Ctrl + T Hotkey Live agent status, token counters, and active provider info Hotkey
Ctrl + C Hotkey Interrupt active streaming response or tool execution Hotkey
/exit or /quit None Exits the thwip terminal multiplexer /exit