Quickstart | ARGUS Docs

How to use ARGUS

Install

pip install argus-agents

Init

argus init — writes .cursor/skills/argus-debug/ and .claude/skills/argus-debug/. Commit them. The skill already contains the setup prompt.

Attach

Ask your editor agent to wire ARGUS. (The skill already contains this AI setup prompt; the homepage copy is just a fallback.) ArgusWatcher.attach(graph)

Run

Same as always. Failures print [argus] in the terminal; clean runs stay silent.

Inspect

argus show last, argus fix <id>, or argus ui

Optional — smarter detection

Run argus key set if you want the LLM judge. Skip it and you still get heuristics.

Prerequisites

Requirements

Python 3.9 or higher
pip (or any Python package manager)
A LangGraph pipeline (or any Python callable to test with)

Installation

bash

pip install argus-agents

This is the full product — the argus CLI, the LangGraph adapter, and the local UI (argus ui). No account, no config files, no cloud: ARGUS runs fully local, runs are stored in .argus/runs/, and heuristic detection works out of the box. LLM-powered features stay optional: run argus key set when you want the semantic judge.

Bring Your Own Key (BYOK)

AI-powered detection (semantic judge, LLM investigator, learned trends) uses your own key from the provider of your choice — OpenAI, Anthropic (Claude), or Google (Gemini). Set it once — it's saved locally and reused every session. No key is fine too: ARGUS falls back to heuristic-only detection.

bash

argus key set                        # OpenAI by default — prompts (hidden), saved to ~/.argus/config.json
argus key set --provider anthropic   # or Anthropic / --provider google for Gemini
# or an env var:  export OPENAI_API_KEY=sk-...  (ANTHROPIC_API_KEY / GEMINI_API_KEY)

argus key use anthropic    # switch active provider   ·   argus doctor  # confirm mode

Setup — Pick Whichever Fits Your Code

One call — attach (recommended)

python

from argus import ArgusWatcher

watcher = ArgusWatcher()
app = watcher.attach(graph)         # StateGraph or already-compiled app
result = app.invoke(initial_state)  # run is persisted automatically
print(watcher.run_id)

Constructor form

python

from argus import ArgusWatcher

watcher = ArgusWatcher(graph)       # uncompiled StateGraph
app = graph.compile()
result = app.invoke(initial_state)  # persisted automatically

Both work. No changes to your node functions.

finalize() is optional

attach() wraps invoke() / ainvoke() so the run is written when the call returns — including cyclic graphs. watcher.finalize() is an optional idempotent flush, not required.

Full Example

Here's a complete example — a simple LangGraph pipeline with ARGUS instrumentation:

pythonexample.py

1from argus import ArgusWatcher
2from langgraph.graph import StateGraph
3
4# 1. Define your graph (your existing code)
5graph = StateGraph(AgentState)
6graph.add_node("agent", call_model)
7graph.add_node("tools", tool_node)
8# ... add edges ...
9
10# 2. Attach ARGUS and run
11watcher = ArgusWatcher()
12app = watcher.attach(graph)
13result = app.invoke(initial_state)
14# persisted automatically

View Results

After your run completes, you can view results in several ways:

bash

# List all runs
argus list

# View the most recent run
argus show last

# View a specific run by ID (or 8-char prefix)
argus show run abc12345

# Launch the web dashboard
argus ui

Video coming soon

Running argus show and viewing results in the terminal

Quick walkthrough of the ARGUS CLI trace viewer

Next Steps

Core Concepts — understand Watchers, Detectors, Traces, and Forensics
Configuration — customize detection sensitivity, enable semantic judging, configure storage
CLI Reference — all available commands and flags