Meet Together Link: A Free CLI That Runs Open Models Like Kimi K3 and GLM 5.3 Inside Claude Code, Codex, and OpenCode
MarkTechPost Asif Razzaq ● Covered by 3 sources
Together AI launched a free CLI that lets agents like Claude Code and Codex use open models instead. It swaps expensive closed-model calls for hosted open ones, with receipts built in.
Based on reporting by MarkTechPost, Asif Razzaq — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Together AI has put out Together Link, a free MIT-licensed command-line tool that is now in beta. It plugs into coding agents developers already use and sends them to open models hosted by Together AI instead of the usual default premium model.
The pitch is blunt: keep your existing tool, change the model, and pay less. Together says engineering teams can spend tens of thousands to millions of dollars a month on closed models. Its answer is a curated set of open options that can handle real coding work, including Kimi K3 and GLM 5.3 for harder tasks, plus GLM 5.3 Flash and DeepSeek V4.1 Flash for everyday ones.
Setup is meant to be painless. On macOS or Linux, a single install command gets the CLI in place, and it only needs a Together API key. The launcher can open supported tools like Claude Code, Claude Desktop, Codex, ChatGPT Desktop, OpenCode, and Pi, or start them directly with commands such as togetherlink claude or togetherlink codex. There are also shortcuts like tclaude.
The product is designed to stay out of the way. According to the docs, there’s no local proxy or daemon; each tool talks straight to Together’s hosted gateway. Temporary configuration is created for each launch and removed when the session ends. Claude Desktop and ChatGPT Desktop use reversible profiles instead, so switching back is just a command away.
Together’s auto router is the part that gives the tool its edge. It looks at the first task in a session and sends quick fixes to cheaper models while reserving stronger models for harder jobs. With an Anthropic API key, it can route between Opus 5.5 and GLM 5.3. Without one, it uses GLM 5.3 and GLM 5.3 Flash. Inside Claude Code, the /model menu maps tiers to open models: Opus becomes Kimi K3, Fable becomes GLM 5.3, Sonnet becomes GLM 5.3 Flash, and Haiku becomes DeepSeek V4.1 Flash.
Billing is also built in. Sessions print token and dollar totals when they end, Claude Code shows estimated spend next to the equivalent Opus cost, and togetherlink usage --last 7d shows gateway-tracked spend across sessions. That is the real product here: not just model swapping, but model swapping with a running tab attached.
My take — AI-written commentary, not fact-checked reporting
This is the kind of unglamorous tooling that actually changes behavior. If open models can sit behind the same buttons people already trust, a lot of “we need the expensive one for everything” thinking suddenly looks lazy. Also, receipts are a lovely insult to waste.
Read more about this at: MarkTechPost