Console

Browser conversations

Run 0 or 0 web to open the account-free browser console under Node or the standalone binary. Choose a workspace folder, connect a provider in Connections, and choose the conversation model and autonomy mode. Use Chat, Findings, Workflows, Plugins and Learning for investigation, evidence review and reusable capabilities.

The interactive terminal UI and readline chat REPL are retired. 0 console, 0 chat, 0 tui and 0 watch without a headless prompt print browser guidance. The CLI still supports scans, source reviews, binary research, automation and MCP integrations. Retained terminal modules supply shared session/settings helpers; they do not provide an interactive terminal entrypoint.

Headless prompts

Run one request and exit:

0 chat --prompt "Review the selected repository" --mode standard
0 console --print "Review the selected repository" --mode standard
0 -p "Review the selected repository" --mode standard
printf '%s' 'Summarize the selected findings' | 0 console --print --mode standard

These paths use the engine’s tool registry and stream ordinary text results. --model, --role, --target, --scope, --finding, --max-tool-calls and --db-path configure the headless run. Run 0 chat --help for current options. A provider connection is independent of target authorization and tool isolation.

Saved sessions

Combine a prompt with a saved session id or unique prefix:

0 console --resume SESSION_ID --print "Continue the investigation" --mode standard
0 console --continue --print "Summarize the evidence" --mode standard

Saved-session model and target are used when available. A bare --resume has no terminal picker; use the browser history or supply a session id.

Non-interactive approval limitations

Headless prompts have no interactive tool-approval surface. Requests requiring human approval cannot be completed through a terminal dialog. Use the browser conversation to review those actions. The default headless autonomy mode is YOLO; when the scope plugin is active it requires configured in-scope entries. Explicit restrictions remain in effect. --mode standard is available for a headless run with the appropriate scope and tool policy.

Execution and models

Local means tools execute on the engine host. Your selected model endpoint can still be external. A VM execution profile and a model connection are separate choices. Configure and inspect them in the browser before starting work.

See Getting started, provider connections, scope and engine connections.

Model picker

Use the browser conversation model picker to select a provider/model. Worker-role model overrides remain an embedding API capability; terminal keyboard shortcuts are retired. See model routing.

Keyboard shortcuts

Terminal keybindings are retired. Browser controls are available with normal keyboard navigation.

Acquiring public source

In YOLO, a standalone public HTTPS repository checkout can acquire source for review without authorizing tests against the hosting service. Ask the chat to acquire the repository, then inspect it in subsequent tool calls. Checkout does not change engagement scope; configured exclusions and private-network restrictions still apply. See scope rules.