Together Link Puts Open Models Inside Claude Code, Codex and Other Coding Agents
Together AI has released Together Link, a tool that connects coding agents teams already use to open models hosted on its platform. According to the company, it can cut spend by over 50% while keeping the same agent and workflow.
- Together Link connects Claude Code, Claude Desktop, Codex, OpenCode and Pi to open models on Together AI.
- Together AI claims it cuts spend by over 50% with the same agent and workflow.
- Setup is one command, and returning to native closed models takes one command.
- The Auto router picks a model once per session, so prompt caching keeps working.
- A per-session tracker compares spend with the cost of the same session on Opus 5.5.
What Together Link does
Together Link links existing coding harnesses to open models served by Together AI. The company says it supports Claude Code, Claude Desktop, Codex (in the ChatGPT app and the CLI), OpenCode, and Pi.
Together AI argues that the cost problem comes from habit: coding agents are now commonplace for engineering teams, and every task, from a one-line fix to a full rewrite, often runs on the same premium model. At scale, the company says, engineering organizations spend anywhere from tens of thousands to millions of dollars a month on closed models.
The pitch rests on the claim that open models have meaningfully closed the gap with top models from closed labs. Together AI names Kimi K3 and GLM 5.3 as models that now handle many of the hardest coding tasks at a fraction of the per-token price. For everyday work, it points to smaller models such as GLM 5.3 Flash and DeepSeek V4.1 Flash.
Setup, routing and cost tracking
Installation is a single command: curl -fsSL https://link.together.ai/install | bash. According to Together AI, the agent then opens exactly as before, now connected to Together AI through the user's Together AI API key. Settings and logins stay as they were, and returning to the native closed models takes one command.
Users can choose the "Auto" option, which relies on the Together AI Router. The router reads the first task of each session and sends it to a suitable model: quick fixes go to fast, low-cost models, while hard problems get frontier capability.
- With an Anthropic key, the router chooses between Opus 5.5 and GLM 5.3.
- Without one, it chooses between GLM 5.3 and GLM 5.3 Flash.
Routing happens once per session, so prompt caching keeps working, the company says. After each session, a tracker shows what the user spent next to what the same session would have cost on Opus 5.5.
Billing runs on the existing Together API key, against serverless pay-as-you-go or credit packs. No separate contract is needed.
Infrastructure and what it means for teams
Together Link runs on Together's serverless inference, which the company describes as the same infrastructure developers already pick for these models on OpenRouter. As of 9/30/2026, Together AI says it serves the largest share of OpenRouter tokens for DeepSeek V4.1 Flash (40.8%), GLM 5.3 Flash (28.2%), and Kimi K3 (23.1%), with competitive speed and pricing across the board.
For engineering teams, the offer is a way to test cheaper open models without changing tools or retraining developers. The savings figure of over 50% is Together AI's own claim, and the per-session tracker is the mechanism meant to let teams check it against their own usage.
Together AI invites teams to create an account to get an API key and to read its documentation. Organizations planning a rollout across an engineering org can contact the company for help.
