Together Link is a free, open-source command line tool that connects coding agents such as Claude Code, Codex and OpenCode to open models on Together AI, with automatic routing and a cost receipt for every session.
Together Link is a free, open-source tool from Together AI that points an existing coding agent at open-weight models hosted on Together's platform. Rather than asking developers to adopt a new editor or a new agent, it sits between the harness they already use and the model provider. Supported harnesses include Claude Code, Claude Desktop, Codex, ChatGPT Desktop, OpenCode and Pi. The product is in beta, and its model lineup covers Kimi K3, GLM 5.3, MiniMax M3 and DeepSeek V4.1 Flash.
In practice, setup is one install command followed by a configure step that stores a Together AI API key. Each agent then launches through its own command, such as togetherlink claude. Terminal agents receive settings that last only for the session, while the desktop apps run on separate, reversible profiles, so logins, history and config files stay where they were and a single off command restores the original setup. The default model choice is Auto Router, which reads the first task of a session and sends quick work to a low-cost model and harder problems to a more capable one. Models can also be picked by hand from the agent's own model menu. When a session ends, a receipt prints token and dollar totals, and a usage report covers the last seven days of spend.
The audience is developers and engineering teams whose agent bills have grown with daily use and who want cheaper models without giving up a familiar workflow. Together states that moving everyday coding work to open models cuts spending by more than half, which is a vendor figure rather than an independent measurement. Compared with wiring up a generic router or swapping provider keys by hand, the pitch here is narrow: keep the harness, add routing and cost visibility. It requires macOS or Linux and a Together AI API key, and usage is billed per token at Together's rates, so it suits people comfortable with terminal commands more than newcomers to coding agents.