Together Link: A Free CLI That Runs Open Models Like Kimi K3 and GLM 5.3 Inside Claude Code, Codex, and OpenCode
By Asif Razzaq | October 5, 2026
Together AI has released Together Link, a free, MIT-licensed CLI now in beta. It connects the coding agents developers already use to open models hosted on Together AI. Supported tools include Claude Code, Claude Desktop, Codex, ChatGPT Desktop, OpenCode, and Pi. The premise is simple: keep the harness, swap the model, and shrink the bill.
Is It Deployable Today?
Yes. Together Link installs with a single command on macOS or Linux and requires only a Together API key. Because it remains in beta, commands, routing behavior, and the supported model list may change.
How Together Link Works
The CLI acts as a lightweight bridge between your existing coding agent and Together AI's hosted open models. Instead of being locked into a single proprietary model per harness, you can point the agent at supported open-weight alternatives.
This matters more in 2026 than it did a year ago. Open-weight frontier models now trade blows with closed counterparts on code generation, long-context reasoning, and agentic tool use, while typically costing a fraction per token. Together Link lets teams capture that price-performance advantage without rewriting prompts, retraining agents, or reworking their local development workflow.
Models and Supported Agents
Together Link is designed for open models such as Kimi K3 and GLM 5.3, served through Together AI's inference infrastructure. Rather than binding to one model, it exposes a configurable routing layer, so you can switch models or route different requests to different endpoints.
On the client side, it plugs into the harnesses developers already rely on:
- Claude Code and Claude Desktop
- Codex and ChatGPT Desktop
- OpenCode
- Pi
The practical effect: teams can keep their agentic workflows, tool definitions, and repo-level context intact, then change which model handles the tokens underneath.
Why It Matters
Three forces make Together Link relevant right now:
- Agent harnesses have matured. Developers rarely want to abandon a harness they have tuned. Decoupling the model from the harness preserves that investment.
- Open models are competitive. With Kimi K3, GLM 5.3, and similar releases, open-weight options are viable defaults rather than fallbacks for cost-sensitive use cases.
- Cost discipline is a first-class concern. Agentic coding burns tokens quickly through long contexts and iterative tool calls. Routing the same workload to cheaper, high-quality open models can materially change the economics of running agents all day.
- Request a Together API key from your Together AI account.
- Install Together Link with the provided one-line command for macOS or Linux.
- Point your preferred agent (Claude Code, Codex, OpenCode, or Pi) at Together Link.
- Select a supported open model such as Kimi K3 or GLM 5.3, and start coding.
Getting Started
Because the project is in beta, expect iteration on the model roster and routing controls. For teams evaluating open-model agent stacks in 2026, the barrier to entry is now roughly one command and an API key.
Add Marktechpost as a preferred source on Google.
via MarkTechPost