togetherlink
Grok Build CLI · GLM 5.2 model · 8 min

Launch Grok Build with GLM 5.2

Keep Grok Build's terminal workflow, local coding tools, and sessions while GLM 5.2 handles model requests through Together AI. Grok Build provides the coding interface; GLM 5.2 produces the answers.

By TogetherLink

TLtogetherlinkDEFAULT-MODEL LAUNCHLaunch Grok Buildwith GLM 5.2$tgrokGLM 5.2 · Chat CompletionsGrok BuildPOWERED BY
The important distinction

Grok Build is the interface. GLM 5.2 is the model.

TogetherLink starts the existing grok binary with a local metadata-only model catalog. The explicit selection points directly at GLM 5.2 on Together AI. No xAI model endpoint is involved in this launch.

Tool
Grok Build
Model
GLM 5.2
Inference
Together AI
Setup

Install both CLIs, then launch

1

Install TogetherLink

$curl -fsSL https://togetherlink.vercel.app/install.sh | bash
2

Install Grok Build if it is not already present

$curl -fsSL https://x.ai/cli/install.sh | bash
3

Save your Together API key

$togetherlink configure

If it asks for an Exa key, press Enter to skip it. Grok Build's native web search is not used in this setup.

4

Launch Grok Build on GLM 5.2

$togetherlink --model zai-org/GLM-5.2 grok

Use the full togetherlink --model ... grok form in scripts when you want the model choice to be explicit.

Your normal Grok configuration stays in place

TogetherLink leaves GROK_HOME alone, so Grok reads your normal settings, built-ins, workflows, plugins, and sessions directly. Only authentication is isolated in a temporary empty path, preventing a saved xAI login from overriding the Together API key. The local catalog endpoint serves model metadata only; inference goes directly to Together AI. Your ~/.grok/config.toml is not rewritten.

Verified output

Read the relevant field from the final JSON event

A real headless run on July 21, 2026 returned the requested text and named GLM 5.2 in the final modelUsage object. This field verifies the backend without asking the model to identify itself.

"modelUsage":{"zai-org/GLM-5.2":{"modelCalls":1}}
Automation

Run one coding task and stream JSON

TogetherLink options come before grok. Grok Build options come after it. The final JSON event includes token usage and the model ID used for the run.

$togetherlink --model zai-org/GLM-5.2 grok --output-format streaming-json -p "Review the current diff"
Common questions

FAQ

Does Grok Build support GLM 5.2?+

Yes. TogetherLink gives Grok Build a local metadata-only model catalog and can select zai-org/GLM-5.2 on Together AI explicitly.

Is the model still Grok?+

No. Grok Build is the command-line coding tool in this setup. GLM 5.2 is the model, and Together AI serves the request.

Do I need an xAI API key?+

No xAI key is used for this setup. You need a Together API key because the selected model runs on Together AI.

Does TogetherLink overwrite my Grok Build config?+

No. It uses your normal Grok home without changing config.toml. A temporary auth path keeps your xAI login separate from the Together launch, while Grok's settings, built-ins, workflows, plugins, and sessions remain available.

Do Grok Build workflows work with TogetherLink?+

Yes. TogetherLink enables workflows for the launch and keeps Grok's native bundled workflow resources available from your normal Grok home.

Can I run Grok Build headlessly with GLM 5.2?+

Yes. Pass Grok Build's prompt and output flags after the grok subcommand. Keep TogetherLink's model flag before the subcommand.

Does Grok Build web search work with TogetherLink?+

No. TogetherLink disables Grok Build's native xAI-backed web search for this launch. Local coding tools and model requests through Together AI still work.

Do Grok Build image generation tools work with TogetherLink?+

Not yet. TogetherLink disables Grok Build's xAI Imagine image generation and editing tools because they use a separate xAI-only API and model. Vision input through supported Together models is unaffected.

Related guides