Together AI Chat Completions Skill for Claude Code

What it is

This is a Claude Code skill that teaches AI coding agents how to correctly call Together AI's chat completions API. It provides the exact SDK patterns, model identifiers, and parameter configurations needed to integrate Together AI into your applications. The skill covers streaming responses, tool calling, JSON mode, and model selection.

The skill targets developers using Claude Code, Cursor, or Codex CLI who need to build applications with Together AI's inference API. Instead of reading API documentation and translating it into code, the agent already knows the correct patterns.

How it saves time or tokens

Without this skill, an AI coding agent guesses at Together AI's API patterns based on its training data, which may be outdated or incorrect. The skill provides verified, current SDK patterns that work on the first try. This eliminates debugging cycles caused by wrong parameter names, deprecated model IDs, or incorrect streaming implementations. The token savings come from avoiding retry prompts and error correction.

How to use

1. Install the skill into your Claude Code environment:

npx skills add togethercomputer/skills
# Or copy to ~/.claude/skills/together-chat-completions/

1. The skill activates automatically when you ask Claude Code to write Together AI integration code.

1. Ask Claude Code to implement Together AI chat completions, and it will use the correct SDK patterns from the skill.

Example

Basic Together AI chat completion with streaming:

import together

client = together.Together(api_key='your-key')

stream = client.chat.completions.create(
    model='meta-llama/Llama-3-70b-chat-hf',
    messages=[{'role': 'user', 'content': 'Explain transformers briefly'}],
    stream=True,
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end='')

The skill ensures the agent uses the correct model identifier format, streaming iteration pattern, and error handling.

Related on TokRepo

• AI Tools for API — API integration tools and patterns for AI services
• AI Tools for Coding — Coding assistants and skills that enhance developer productivity

Common pitfalls

• Together AI model identifiers use the format 'org/model-name'. Using short names without the organization prefix causes 404 errors.
• Streaming responses require iterating over chunks and checking delta.content for None values. Missing this check causes TypeError exceptions.
• The Together AI SDK is a separate package from OpenAI's SDK. Installing together instead of openai is required even though the API is OpenAI-compatible.
• Always check the official documentation for the latest version-specific changes and migration guides before upgrading in production environments.
• For team deployments, establish clear guidelines on configuration and usage patterns to ensure consistency across developers.