Reference/CLI Reference

Reference

CLI Reference

Complete reference for all ARGUS CLI commands.

Overview

The ARGUS CLI is your interface for viewing runs, replaying executions, generating reports, and managing your workspace. Install it with pip install argus-agents — the CLI ships with the Python package.

bash
argus init                           # write Cursor + Claude project skills
argus list                           # all runs
argus show last                      # most recent run
argus show run <id>                  # by full id or 8-char prefix
argus fix <id>                       # paste-ready prompt for the root-cause node
argus replay <id> <node>             # re-run from a node
argus replay <id> <node> --only      # re-run just that one node
argus inspect <id> --step <node>     # raw input/output for a node
argus diff <id>                      # rerun vs original
argus diff <id-a> <id-b>             # any two runs
argus ui                             # open web dashboard
argus doctor                         # check setup + LLM mode (BYOK/hosted/heuristic)
argus key set [--provider ...]       # save a provider key locally — OpenAI/Anthropic/Google (BYOK)
argus key use <provider>             # switch the active provider
argus key show                       # list configured providers (masked); * marks active
argus key clear [--provider ...]     # remove one provider's key, or all
argus login                          # (optional) sign in for hosted cloud sync
argus logout                         # clear credentials
argus whoami                         # check login status
argus update                         # check for new release

argus init

Write Cursor and Claude project skills (.cursor/skills/argus-debug/ and .claude/skills/argus-debug/). Commit them. The skill already contains the setup prompt — later chats can call ArgusWatcher.attach() and read .argus/runs instead of guessing from logs. Safe to re-run; pass --force to overwrite customized files.

bash
argus init
argus init --force

argus list

List all stored runs.

bash
# List all runs
argus list

argus fix

Print a paste-ready prompt for a coding agent, aimed at the root-cause node — not the crash site. No LLM call. Write to a file with --output. Strip recorded values with --sanitized.

bash
argus fix <run-id>
argus fix <run-id> --node retrieve
argus fix <run-id> --output fix.md
argus fix <run-id> --sanitized

argus show

View a specific run in your terminal. Shows the full trace with node statuses, timing, detections, and root cause analysis.

bash
# View the most recent run
argus show last

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

Example output:

text
argus  run-abc12345  ·  2024-04-05 12:30  ·  1243 ms
status  ●  silent_failure

   1  fetch       43 ms    ✓  pass
   2  validate    12 ms    ⚠  silent failure
      └─  Field "score" is missing
      └─  process received bad state
   3  process    891 ms    ✗  crashed
      └─  KeyError: 'score'
      └─  Field 'score' was absent from the incoming state

root cause   validate

Node Statuses

  • ✓ — pass
  • ~ — pass with warnings (empty optional fields)
  • ⚠ — silent failure (missing required fields)
  • ⊗ — semantic fail (validator returned False)
  • ⏸ — interrupted (human-in-the-loop pause)
  • ○ — skipped (conditional branch not taken)
  • ✗ — crashed

argus inspect

View the raw input/output state for a specific node in a run.

bash
# Inspect a specific node's raw data
argus inspect <id> --step <node>

argus replay

Re-execute a pipeline from a specific node. ARGUS restores the exact state at that node from disk and runs from there. Upstream outputs stay frozen — only the target node onward re-executes with your fixed code.

bash
# Replay from a specific node
argus replay <run-id> node_7

# Re-run just one node in isolation
argus replay <run-id> node_7 --only

All external HTTP calls (OpenAI, search tools, databases) are recorded by default. During replay, the recorded responses are served back — same data, zero extra cost, fully reproducible.

Replay requires persist_state

Replay only works on runs recorded with persist_state=True (the default). If state persistence was disabled, ARGUS doesn't have the intermediate states needed to replay.

argus diff

Compare two runs side by side. When a replay finishes, ARGUS automatically compares it against the original using an LLM — showing per-node diffs of what changed, what improved, and whether the fix actually worked.

bash
# Compare a rerun against its original
argus diff <rerun-id>

# Compare any two runs
argus diff <id-a> <id-b>

argus ui

Launch the local web dashboard in your browser. Serves runs from .argus/runs/ in your current directory — no account needed.

bash
# Launch the dashboard
argus ui

Opens at http://localhost:7842. The UI includes:

  • Runs List — all stored runs with status, timing, and detection counts
  • Run Detail — node-by-node trace with inputs, outputs, and detections
  • Compare Runs — side-by-side diff between any two runs
  • Approvals — review and approve AI-discovered failure signatures
  • Report Board — submit diagnostic reports for bugs or issues
  • Settings — configure Linear integration and cloud sync

From the Run Detail page, hover any step and click ↺ Rerun From Here to trigger a replay. After rerun, the diff view opens automatically.

argus doctor

Diagnose setup issues. Checks Python version, LangGraph compatibility, storage health, rerun readiness, and optional dependencies.

bash
argus doctor
text
✓  python           Python 3.9.6
✓  langgraph        langgraph 0.6.11
✓  storage          312 runs stored, all healthy
✓  replay           all 7 node functions importable for rerun
✓  optional deps    openai (key set), dotenv

argus key set / show / clear

Bring your own key (BYOK). Pick your provider — OpenAI, Anthropic (Claude), or Google (Gemini) — save the key once and it's reused every session. AI-powered detection (semantic judge, LLM investigator, learned trends) runs entirely on your key, fully local; ARGUS picks a sensible model per call. Per-provider resolution order: env var ( OPENAI_API_KEY / ANTHROPIC_API_KEY / GEMINI_API_KEY) → saved key → hosted proxy (cloud tier) → heuristic-only.

bash
# Save a key (prompts hidden, stored at ~/.argus/config.json). OpenAI by default:
argus key set
argus key set --provider anthropic      # or Anthropic (Claude)
argus key set --provider google         # or Google (Gemini)

# Or pass it directly / use an env var
argus key set sk-... --provider openai
export OPENAI_API_KEY=sk-...            # or ANTHROPIC_API_KEY / GEMINI_API_KEY

# Switch, inspect, remove
argus key use anthropic    # activate a provider you already configured
argus key show             # list configured providers (masked); * marks active
argus key clear            # remove all (or --provider <name> for one)

argus login / logout / whoami

Optional — hosted/enterprise tier only. Authentication for hosted cloud sync and the shared trends registry. The open-source package is fully local and needs no login; argus login reports a hosted-only message unless a hosted backend is configured.

bash
# Sign in for cloud sync
argus login

# Clear credentials
argus logout

# Check login status
argus whoami

argus update

Check for and install ARGUS updates.

bash
# Check for a new release
argus update
Previous

Adaptive Learning

Next

API Reference