argus-hub
Since 0.6Changed in next
The Talos Hub: installs, updates and opens Talos engines
Usage
argus-hub [OPTIONS] [-- <AGENT ARGS>...]
Arguments
| Argument | Description |
|---|---|
[AGENT ARGS]... | With --mcp-agent: more arguments for argus-agent, after -- |
Options
| Flag | Value | Default | Allowed values | Description | |
|---|---|---|---|---|---|
--open | <PROJECT> | Open a project (a folder or its argus.project.json) with the engine it names --set-engine <PROJECT> <SERIES> Change the engine series a project opens with: --set-engine <PROJECT> <SERIES> --upgrade-report <PROJECT> [SERIES] What upgrading a project to a newer installed series would change: --upgrade-report <PROJECT> [SERIES] | |||
--allow-older | With --set-engine: allow a series older than the one the project was saved with | ||||
--import-pack | <PACK> | Install an engine from an offline .argus-pack, with no network | |||
--verify | [<VERSION>] | Re-hash an installed version, or every one | |||
--repair | <VERSION> | Re-fetch the files of an installed version that fail their check | |||
--install | `[<SERIES | VERSION>]` | Install a series ("0.5"), a version ("0.5.1"), or with nothing named, what is recommended | ||
--install-sample | <ID> | Install a sample by id from the update source's samples index | |||
--plugins | List the plugins installed here and the ones the update source offers | ||||
--install-plugin | <ID[@VERSION]> | Install a plugin from the update source: <ID> or <ID>@<VERSION> | |||
--uninstall-plugin | <ID[@VERSION]> | Remove an installed plugin: <ID>@<VERSION>, or <ID> for every installed version | |||
--dest | <DIR> | With --install-sample: where the sample goes (default <projects folder>\Samples\<id>) | |||
--offline-dir | <DIR> | With --install-sample: a folder searched first for a recipe's fetch files (repeatable) | |||
--uninstall | <VERSION> | ||||
--check | Ask the update source what it offers | ||||
--info | |||||
--token-stdin | Read an access token from standard input and store it | ||||
--login | Sign in with a code shown here and approved in a browser (opens nothing itself) | ||||
--logout | End this device's sign-in on the server and remove the stored token | ||||
--send-reports | Send the bug reports waiting in the outbox | ||||
--connect-ai | <PROJECT> | Connect your own AI assistant (Claude Code, or any MCP client) to a project: writes its .mcp.json with the argus-agent of the installed engine that opens it | |||
--client | <CLIENT> | With --connect-ai: which kind of AI client starts the server, windows (the default) or wsl (Claude Code running inside WSL: the command is the launcher's /mnt/<drive>/ path). A project named as /mnt/<drive>/... means wsl | |||
--approve | With --connect-ai: also approve the server in the project's .claude/settings.local.json, so Claude Code loads it without asking | ||||
--allow-tools | With --connect-ai: also let Claude Code use the server's tools without asking, by adding them to permissions.allow in the project's .claude/settings.local.json. Off by default | ||||
--mcp-agent | <PROJECT> | Run the argus-agent of the installed engine that opens this project, on this process's standard input and output (what --connect-ai's .mcp.json starts) | |||
--source | <URL> | Use this update source for this run | |||
--channel | <CHANNEL> | Use this update channel for this run ("nightly" for internal testers); Settings keeps its own | |||
--set-channel | <CHANNEL> | Save the update channel in Settings (the default channel, stable unless the build says otherwise, puts it back) | |||
--quiet | Never show the window; print only failures | ||||
--json | With --info, --check, --install or --import-pack: one JSON object per result per line | ||||
--timeout | <SECONDS> | Stop after this many seconds (exit 124) |
Example
argus-hub [-- <AGENT ARGS>...]