Dev.to AI 🤖 Ai 👁 0 📖 2 min read

Grok Build CLI: x_search Not Working? The Fix Was Call Flags, Not Your Plan

TL;DR: After installing Grok Build (the grok CLI), X search (x_search) didn't work for several days. The cause wasn't my plan, billing, or account permissions — it was missing flags in the invocation. Adding --no-subagen

TL;DR: After installing Grok Build (the grok CLI), X search (x_search) didn't work for several days. The cause wasn't my plan, billing, or account permissions — it was missing flags in the invocation. Adding --no-subagents --max-turns 6 --reasoning-effort low made it work on the spot.

Verified on November 3, 2026.

What happened

Right after setting up Grok Build CLI, I couldn't get X search to run at all. My first suspects were the usual ones: maybe my paid plan hadn't propagated yet, or maybe the account lacked some permission. So I waited, retried, waited some more. Nothing changed.

That assumption cost me days. The search feature itself was fine the whole time.

The actual cause

The fix came from a publicly shared skill called "muse-grok-x-search" (published by Kagami Kami on X). It documented an invocation format I hadn't been using. I copied that format exactly, and x_search worked immediately.

The three flags were:

  • --no-subagents
  • --max-turns 6
  • --reasoning-effort low

In other words, the problem was never the contract or the account. It was how the CLI was being called.

The lesson: check the invocation before the plan

When a tool "doesn't work," the order in which you suspect things matters. I burned several days on plan propagation and permissions, when the only thing missing was a set of flags.

My rule now: before questioning subscriptions, entitlements, or account state, verify the invocation itself — flags, arguments, and required parameters. CLI tools often fail silently or vaguely when called in a shape they don't accept, and that failure looks exactly like an account problem.

How I use x_search, operationally

One more note on usage: I treat X search results as a way to read public sentiment and reputation — what people are saying, how something is being received. I don't treat them as proof of individual facts. Anything factual gets cross-checked against a separate source before I rely on it.

Caveats / what I did not verify

  • I did not isolate which of the three flags was strictly required. What I verified is that the three-flag set works; removing them one at a time to find the minimum is still untested.
  • CLI behavior may change with versions or spec updates. If this doesn't reproduce for you, check the current version's expected invocation format first.
  • This reflects one environment's hands-on result, not an official statement from xAI.

By Fujiko P. hD

📰 Read the original article on Dev.to AI

Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.