Prompt Tags

Overview

Prompt Tags let you label and organize prompt versions with meaningful names like production, staging, or experiment-v2. Tags work everywhere — the platform UI, TypeScript and Python SDKs, the CLI, and the MCP server. Fetch a tagged prompt with the shorthand prompt:tag syntax, and manage tag definitions through a dedicated API.

What You Can Do Now

• Tag prompts in the platform — assign tags to any prompt version from the Prompt Playground.

• Fetch by tag in code — use the prompt:tag shorthand syntax to load a specific tagged version at runtime in both TypeScript and Python SDKs.

• Manage tags from the CLI — create, list, rename, and assign prompt tags with the new CLI commands.

• Tag prompts via MCP — manage tags from Claude Code or Cursor without leaving your editor.

• Load tagged prompts in Playground — load a tagged prompt directly from a trace reference to see exactly what was running in production.

• Custom tag definitions — create your own tag names with the tag definitions API, beyond the built-in defaults.

Rollout

Available now across all LangWatch surfaces — platform, TypeScript SDK, Python SDK, CLI, and MCP.

Notes

• Tags are managed per-organization via the /api/prompts/tags endpoint.

• The TypeScript SDK supports tag CRUD operations: create, list, fetch, and assign.

• The Python SDK supports custom tag assignment on prompt capture.

• Tags are Prisma-backed with proper relational integrity (tagId foreign key).