Revolutionizing Developer Workflows with Open-Source AI
Together AI has officially unveiled Together Link, a new, free command-line interface (CLI) currently available in beta, under an MIT license. This tool is engineered to bridge existing coding environments, such as Claude Code, Claude Desktop, Codex, and OpenCode, with a suite of high-performance open models hosted by Together AI. Developers can thus retain their existing agent workflows while seamlessly accessing more cost-effective AI models.
The CLI supports macOS and Linux, featuring a swift, one-command installation that requires only a Together AI API key. While in its beta phase, users should anticipate potential adjustments to commands, routing mechanisms, and the roster of available models.
Addressing the High Cost of AI-Powered Development
The development community frequently encounters a challenge where coding agents default to premium, closed-source models for every task, irrespective of complexity. These operations can incur substantial monthly expenses for engineering organizations, often reaching millions on proprietary models. Together AI argues that open-source models have significantly advanced, now effectively matching or exceeding the capabilities of their closed counterparts for many critical programming tasks. For instance, Kimi K3 and GLM 5.3 are tailored for demanding coding work, while GLM 5.3 Flash and DeepSeek V4.1 Flash proficiently handle routine development activities.
Seamless Integration and Intelligent Routing
Implementing Together Link is designed for simplicity. A single curl command initiates the installation, which also installs Bun if necessary, placing the executables in the user's local binary directory. Users can launch a unified interface or directly invoke specific tools, such as togetherlink claude or togetherlink codex. Convenience shortcuts, like tclaude, are also supported.
The architecture of Together Link is notable for its efficiency: it operates without requiring a local proxy or daemon. Instead, each integrated tool communicates directly with Together AI's hosted gateway. For terminal-based agents, temporary configurations are applied at launch and removed upon session completion. For desktop agents like Claude and ChatGPT Desktop, separate, reversible profiles are employed.
A key feature is the "Auto Router," which intelligently processes the initial task of each session to select the most suitable model. Simpler fixes are routed to fast, economical models, while complex challenges are directed to more advanced capabilities. For users with an Anthropic API key, the router can even arbitrate between Anthropic's Opus 5.5 and GLM 5.3. Without an Anthropic key, it intelligently routes between GLM 5.3 and GLM 5.3 Flash. This routing decision is made once per session, ensuring efficient prompt caching.
Within Claude Code, the /model menu dynamically maps tiers to specific open models: Opus is linked to Kimi K3, Fable to GLM 5.3, Sonnet to GLM 5.3 Flash, and Haiku to DeepSeek V4.1 Flash. Users can also pin a specific model using a flag before the tool name.
Transparent Pricing and Cost Management
Together Link leverages an existing Together AI API key, operating on a pay-as-you-go or credit pack basis. Supported models include Kimi K3, GLM 5.3, GLM 5.3 Flash, and DeepSeek V4.1 Flash, each offering a substantial 1M context window. Pricing is competitive, with Kimi K3 listed at $3.00 in and $15.00 out, and GLM 5.3 at $1.40 in and $4.40 out. Up-to-date rates are available on Together AI's official pricing page.
For financial oversight, each session concludes with a summary of token usage and associated dollar costs. In Claude Code, an estimated spend is displayed directly, alongside a comparison to the equivalent Opus cost. A togetherlink usage --last 7d command offers a comprehensive report of gateway-tracked spending across all sessions.
Distinguishing Features and Advantages
Together Link differentiates itself from alternatives through several key aspects. Its primary advantage lies in a streamlined, single-command setup combined with integrated cost receipts, offering a clear path to savings. Unlike solutions that might require managing a local gateway, Together Link operates entirely via a hosted gateway, simplifying infrastructure. While alternatives may offer broader provider control, Together Link provides a curated selection of high-performance models with built-in financial transparency, making it a compelling option for developers prioritizing efficiency and cost reduction.
- One simple installation command connects six existing agents to open models on Together AI.
- The default "Auto Router" intelligently selects between models like GLM 5.3, GLM 5.3 Flash, and optionally Opus 5.5.
- Reported cost savings exceed 50%, potentially reaching 50 to 80% compared to exclusive Opus 5.5 usage.
- No local proxy is required, preserving original agent configurations.
- The tool is free and MIT-licensed, currently available for macOS and Linux during its beta development phase.
This article is a rewritten summary based on publicly available reporting. For the original story, visit the source.
Source: MarkTechPost
