Argus

No test selected

Select a test on the left, then Run.

No tests? Click init to scaffold .argus/ with an example.

Free roam

Let the model explore the application like a curious child — it pokes at everything, documents what it does, and reports the bugs it finds.

Target command

The application to launch and explore.

Time budget (minutes)

Token budget (optional)

Ollama is local & free — time budget only.

status idle findings 0 tokens 0

No session yet.

Provider

Argus is provider-agnostic: Ollama (local), Anthropic, OpenAI, Azure, Gemini, LiteLLM. Configure in .argus/config.yaml.

provider … model … time budget … token budget …

Token usage

this session 0 project total 0 LLM calls 0

Knowledge Engine

Argus builds a semantic graph of UI states and transitions across sessions. Past bugs and unexplored paths are injected into the LLM context — every run is smarter than the last.

Target

—

states

—

transitions

—

bugs found

—

sessions

How it works

State graph

Each observed UI state is fingerprinted and stored as a node. Actions become directed edges. Red nodes mark known bug zones.

Vector recall

Observations are embedded with all-MiniLM-L6-v2. Similar past states and bug reports are retrieved and injected into the LLM prompt.

Cross-session

Knowledge persists in .argus/knowledge/. Configure knowledge: in .argus/config.yaml to switch backends.