Blackbox AI adds /auto command
Blackbox AI launched a new /auto command that delegates tasks automatically to top agents based on developer usage data, aiming to streamline multi‑agent workflows for coding tasks. The feature was presented as using 30M+ dev usage signals to select agents like Claude Code and Codex CLI for optimal outcomes. The announcement reflects ongoing moves to orchestrate agents rather than rely solely on single‑prompt interactions. (x.com)
Blackbox AI has added a new `/auto` command to its coding tool, pitching it as a way to route a task to the agent most likely to handle it well. (blackbox.ai) Blackbox’s command-line tool already lets developers run several coding agents from one terminal, including Blackbox, Claude Code, Codex, Gemini, Goose, OpenCode, and Qwen. In its existing `/multi-agent` mode, each agent works on a separate Git branch and a “Chairman” model picks the best implementation. (blackbox.ai) (docs.blackbox.ai) The company presented `/auto` in an April 2026 post on X, saying the command uses more than 30 million developer usage signals to decide which agent to call for a given coding task. The demo named Claude Code and Codex Command Line Interface as examples of agents the system can dispatch. (x.com) That setup changes the developer’s job from choosing a model by hand to describing the task and letting the software pick a specialist. Blackbox’s own product pages already frame the tool as one terminal that can “run blackbox, claude code, codex, gemini” and compare outputs across models. (blackbox.ai) The underlying idea is simple: instead of asking one model to do everything, the tool can send work to different agents that have different strengths. Blackbox’s documentation says multi-agent runs create separate branches, execute in parallel, and then compare the results. (docs.blackbox.ai) Blackbox has been building toward this structure for months. Its onboarding flow says users can “dispatch competing agents from a single command,” and its Command Line Interface page says the product can open pull requests after agents analyze a codebase in parallel. (blackbox.ai 1) (blackbox.ai 2) The company is also tying that orchestration pitch to scale. Blackbox’s site says 30 million developers use its products, and the `/auto` announcement turns that figure into a ranking signal for deciding which agent gets a task. (blackbox.ai) (x.com) What Blackbox has not published in its public documentation is a detailed breakdown of how those usage signals are weighted, when `/auto` overrides a user’s default agent, or whether developers can inspect the routing decision after a task runs. Its public command reference currently documents `/agent`, `/multi-agent`, `/skill`, and `/tasks`, but not `/auto`. (docs.blackbox.ai) For now, `/auto` looks like Blackbox’s next step in turning the coding assistant from a single chatbot into a traffic controller for several agent systems inside one terminal. (blackbox.ai)
