Together Link connects coding agents to open models with one command

Together AI has released Together Link, a tool that connects existing coding agents to open models on its serverless platform. The setup runs from a single command, and the agent opens as before, now using a Together AI API key. Supported harnesses include Claude Code, Claude Desktop, Codex (in the ChatGPT app and CLI), OpenCode, and Pi. Settings and logins stay unchanged, and switching back to native closed models takes one command.

Router picks models per session

Together Link includes an "Auto" mode that uses the Together AI Router. It reads the first task of each session and sends it to an appropriate model: quick fixes go to fast, low-cost models, and harder problems go to frontier-capable ones. If the user brings an Anthropic key, the router chooses between Opus 5.5 and GLM 5.3. Without that key, it routes between GLM 5.3 and GLM 5.3 Flash. Routing happens once per session, so prompt caching continues to work.

A per-session tracker shows what the session cost and what the same session would have cost on Opus 5.5. Billing runs on the existing Together API key, using serverless pay-as-you-go or credit packs, with no separate contract.

Cost context for engineering teams

The company states that engineering organizations spend from tens of thousands to millions of dollars per month on closed models. Open models like Kimi K3 and GLM 5.3 now handle many hard coding tasks at a fraction of the per-token price, while smaller models such as GLM 5.3 Flash and DeepSeek V4.1 Flash cover everyday coding. Together AI claims that Link cuts spend by over 50% for the same agent and workflow.

Together Link runs on Together's serverless inference, the same infrastructure used for these models on OpenRouter. According to Together AI, as of September 30, 2026, 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%). The company cites competitive speed and pricing across the board.

For inference engineers, the key point is that routing is session-level, not per-request, which preserves prompt caching. Link also avoids vendor lock-in at the harness level: teams keep their existing agent and can revert to closed models at any time.

Together AI provides documentation and offers planning help for org-wide rollouts. The tool is available now through the install command and a Together AI account.