Open Source · Apache 2.0

Every OpenRouter model.
One terminal.

Query GPT-4o, Claude 3.5, Mistral, Llama, Gemini, and 200+ more — with streaming markdown responses, live model switching, and per-message cost tracking.

$ pip install switchboard-tui

View on GitHub →

Switchboard claude-3-5-sonnet · 128K ctx

Write a Python function to debounce async calls

Here's a clean async debounce implementation:

asyncio.create_task cancels the previous task on each call, so only the last invocation within the delay window actually runs.

```python
import asyncio
from typing import Callable, Any

def debounce(delay: float):
    task = None
    async def wrapper(fn, *args):
        nonlocal task
        if task: task.cancel()
        ...

Ask anything...

Send

in: 18 tok out: 143 tok cost: $0.002295 model: anthropic/claude-3-5-sonnet

Features

Model browser

Ctrl+K opens a searchable list of every OpenRouter model with provider, context length, and pricing. Filter by name or provider in real time.

Streaming markdown

Responses stream token by token and render as markdown — code blocks, lists, and formatting update live as they arrive.

Switch mid-conversation

Change models at any point without losing history. Compare how different models handle the same context.

Cost tracking

Every message shows tokens in, tokens out, and the estimated USD cost. See exactly what each exchange costs across models.

System prompt editor

Ctrl+S opens a full editor for the system prompt. Change persona, context, or constraints without restarting the session.

Free models highlighted

OpenRouter's free-tier models are clearly marked in the browser. Start experimenting at zero cost before switching to paid models.

Sample models

Model	Provider	Context	In $/M	Out $/M
claude-3-5-sonnet	anthropic	128K	3.00	15.00
gpt-4o	openai	128K	5.00	15.00
gemini-flash-1.5	google	1M	0.075	0.30
mistral-7b-instruct:free	mistralai	32K	—	—	FREE
llama-3.1-8b-instruct:free	meta-llama	128K	—	—	FREE
deepseek-chat	deepseek	64K	0.14	0.28

200+ models available via openrouter.ai/models. Free-tier models need no credits.

Quickstart

Install

pip install switchboard-tui

Get an API key

export OPENROUTER_API_KEY=sk-or-...

Launch

            # Start with the default model

            switchboard

            # Or specify a model directly

            switchboard --model openai/gpt-4o

Optional config

            # ~/.config/switchboard/config.toml

            api_key = "sk-or-..."

            [chat]

            default_model = "anthropic/claude-3-5-haiku"

            temperature = 0.7

Every OpenRouter model.One terminal.