SEER — Shell Enhanced Execution & Reasoning

An AI assistant that lives in your terminal. It sees what just went wrong and tells you how to fix it. It answers shell questions. In brave mode, it does the job for you.

Why seer

• No copy-pasting errors. seer help reads your recent terminal output and explains the last failure.
• It can do the work. Brave mode runs the commands it needs, reads the results, and answers: tables, summaries, fixes.
• Safe by default. Read-only commands run on their own. Anything that changes something asks first, and you see the whole command.
• Your choice of model. Local (LM Studio, Ollama, llama.cpp, vLLM), OpenAI, Anthropic, or your Claude Code / Codex subscription. seer picks up a running local server automatically.
• Plain English at the prompt. Type what you want and press Ctrl+G.

macOS and Linux · zsh and bash · inspired by PEEL for PowerShell.

Quick start

1. Install

curl -fsSL https://raw.githubusercontent.com/zeddius1983/seer/main/install.sh | bash
source ~/.zshrc

This installs seer (and uv if needed), the zsh integration and Ctrl+G. On bash, add source "$(seer --shell-path bash)" to ~/.bashrc instead.

2. Connect a model. Start LM Studio or Ollama with any model and seer finds it. For something else, see Choosing a model. Check with:

seer --stats

3. Use it

seer help                          # after something fails: what went wrong, and the fix
seer how do I find what is using port 8080
seer -b what is taking up space in my home folder

Brave mode: let seer do it

Add -b and seer works out the commands, runs them, and answers with the result. Each command is shown as it runs.

It handles tasks that need several steps, where each step depends on what the last one found:

seer -b find the 5 largest files here and show them as a table
seer -b which file changed most in the last 20 commits, and what were its last 3 changes about
seer -b which of my python files has the most lines, and what does it import
seer -b what is listening on network ports, and which process owns each
seer -b find log files older than 30 days, tell me how much space they take, and compress them
seer -b check whether config.yaml is valid, explain what is wrong, and fix it
seer -b write a zfs cheatsheet and save it to ~/Documents as markdown

Anything that changes something asks first. You see the whole command in a box and answer y, n, or e to edit it. If you answer n, seer stops and tells you what it would have run.

To make brave mode the default for every question, including Ctrl+G, add this to ~/.config/seer/config.yaml:

brave: true
brave_confirm: auto   # when to ask first: auto | trust | always

Everyday use

Useful flags: -s streams the answer as it's written · -r gives plain text for scripts · -p <provider> / -m <model> switch the model for one question · --no-context skips your terminal output.

Diagnostics: seer --stats shows the active model and settings, seer --context shows exactly what gets sent, and seer config prints your config file.

Choosing a model

With provider: auto (the default), seer uses the first local server it finds: llama.cpp, LM Studio, Ollama, then vLLM. To use something else, set it in ~/.config/seer/config.yaml (run seer config to create it):

provider: anthropic        # or: openai, ollama, lmstudio, claude-cli, …

provider: claude-cli     # use it always…
auto_cli: true           # …or only when no local server is running (claude-cli, then codex-cli)

providers:
  claude-cli:
    type: claude-cli
    model: sonnet        # haiku / sonnet / opus, or a full model id
  codex-cli:
    type: codex-cli
    model: [gpt-6-luna, auto]   # tried in order; auto = your ~/.codex/config.toml model
    reasoning_effort: low

provider: myserver
providers:
  myserver:
    type: openai                     # any OpenAI-compatible API
    base_url: http://myserver:8080/v1
    api_key: none
    model: my-model                  # or auto: the first model the server lists

How it works

A small shell hook saves your recent terminal output after every command. seer help sends that output, plus your OS and shell, to the model. Inside tmux seer reads the whole screen, including error output. Outside tmux it sees the last command and its exit code.

curl -fsSL https://raw.githubusercontent.com/zeddius1983/seer/main/install.sh | bash -s -- --version v1.2.0

curl -fsSL https://raw.githubusercontent.com/zeddius1983/seer/main/install.sh | SEER_IMPLICIT_BIND='^@' bash

curl -fsSL https://raw.githubusercontent.com/zeddius1983/seer/main/install.sh | bash -s -- --uninstall

uv sync                  # dev environment
uv run pytest            # tests
.venv/bin/seer --help    # run without installing