Spegel - Reflect the web through AI

January 20, 2026 · View on GitHub

Automatically rewrites the websites into markdown optimised for viewing in the terminal. Read intro blog post here

This is a proof-of-concept, bugs are to be expected but feel free to raise an issue or pull request.

Screenshot

Sometimes you don't want to read through someone's life story just to get to a recipe Recipe Example

Installation

Requires Python 3.11+

$ pip install spegel

or clone the repo and install it in editable mode

# Clone and enter the directory
$ git clone https://github.com/simedw/spegel.git
$ cd spegel

# Install dependencies and the CLI
$ pip install -e .

API Keys

Spegel is using litellm, which allows the use of the common LLMs, both local and external.

By default Gemini 2.5 Flash Lite is used, which requires you to set the GEMINI_API_KEY, see env_example.txt

Usage

Launch the browser

spegel                # Start with welcome screen
spegel bbc.com        # Open a URL immediately

Or, equivalently:

python -m spegel      # Start with welcome screen
python -m spegel bbc.com

Basic controls

  • / – Open URL input
  • Tab/Shift+Tab – Cycle links
  • Enter – Open selected link
  • e – Edit LLM prompt for current view
  • b – Go back
  • q – Quit

Editing settings

Spegel loads settings from a TOML config file. You can customize views, prompts, and UI options.

Config file search order:

  1. ./.spegel.toml (current directory)
  2. ~/.spegel.toml
  3. ~/.config/spegel/config.toml

To edit settings:

  1. Copy the example config: 
    cp example_config.toml .spegel.toml
# or create ~/.spegel.toml
  2. Edit .spegel.toml in your favorite editor.

Example snippet:

[settings]
default_view = "terminal"
app_title = "Spegel"

[ai]
default_model="gpt-5-mini"

[[views]]
id = "raw"
name = "Raw View"
hotkey = "1"
order  = "1"
prompt = ""

[[views]]
id = "terminal"
name = "Terminal"
hotkey = "2"
order = "2"
prompt = "Transform this webpage into the perfect terminal browsing experience! ..."
model="claude-4-5-haiku-20250414"

Local Models with Ollama

To run with a local model using Ollama, first pull and serve your desired model:

$ ollama pull llama2
$ ollama serve

Then set the model in .spegel.toml as follows:

model = "ollama/llama2"

Ollama supports models like Llama, Mistral, and many others.

License

MIT License - see LICENSE file for details.

For more, see the code or open an issue!