LangChain Prompts

You can access LangSmith’s prompt library through this MCP Server to discover, compare, and use community-vetted prompts in your AI workflows. It provides rich metadata, version history, and advanced search to help you find prompts tailored to your use case, with seamless integration across multiple MCP-compatible clients.

How to use

You connect to the LangChain Prompts MCP Server from your MCP client using the provided HTTP endpoint. Once connected, you can search the prompt library, view detailed prompt information, compare prompts side-by-side, validate prompts for quality, and fetch intelligent completions for templates. Use these practical workflows to discover prompts for code reviews, documentation, assistant enhancements, or research.

How to install

Prerequisites: Node.js (for JavaScript/TypeScript tooling) or your chosen MCP client that supports HTTP MCP endpoints.

Step 1: Ensure you have access to the LangChain Prompts MCP Server via its HTTP endpoint.

Step 2: Add the MCP server to your client configuration using the HTTP transport URL https://mcp.prompts.mcpcentral.io/mcp.

Step 3: If your client requires authentication for private prompts, set your LangSmith API key as an environment variable named LANGSMITH_API_KEY in the client configuration or your environment.

Configuration and access notes

Public prompts are accessible without authentication. For access to private prompts and collections, provide your LangSmith API key where required by your MCP client. Prompts include metadata such as downloads, views, likes, tags, and version history. You can filter by owner, description, or tags, and view real-time collections like Popular, Recent, and Trending.

If you use a Claude Desktop or Cursor/Continue.dev client, configure the server as shown in the connection examples. The HTTP endpoint is consistent across clients, and you can opt in to private content by supplying your API key.

Examples and workflows

Example: Discovery and evaluation of code-review prompts. Use advanced search to filter by domain and tag, fetch prompt content, and compare results to choose the best fit for your project.

Example: Building a team prompt library. Find trending prompts, validate their quality, and track version history to standardize usage across your organization.

Example: Enhancing an AI assistant. Analyze current prompts, locate similar high-performing prompts, and apply improvement suggestions to your templates.

Troubleshooting

If you encounter connectivity issues, verify you can reach the endpoint https://mcp.prompts.mcpcentral.io/mcp from your network. Check that your API key (if used) is valid and that your MCP client is configured to use the HTTP transport with the correct URL.

Architecture and capabilities overview

This MCP Server provides a dual-transport design (HTTP for production and stdio for local development), full TypeScript implementation with runtime validation, and intelligent caching. It includes built-in rate limiting and security-focused input handling. You gain access to a broad set of tools for prompts, dynamic resources, and guided prompts to help you discover, analyze, and improve prompts.

Data sources and prompts overview

Primary data comes from LangSmith Prompt Hub, offering 1000+ community prompts with metadata, parameterized templates, and real-time updates. Public prompts are accessible without auth, while private prompts require an API key. You can fetch prompt content, versions, and statistics to understand usage and evolution.

Versioning and compatibility

The server follows MCP 2025-06-18 specifications, with structured outputs, prompt templates, and robust transport options. It is designed to be cloud-optimized and enterprise-ready, with considerations for security and scale.

Testing and discovery

You can exercise the MCP server interactively in the lab environment at the MCP Central Lab endpoint. Use it to test your client integration and validate the end-to-end flow from search to prompt content retrieval.

Support and health checks

For health monitoring, perform a GET request to https://mcp.prompts.mcpcentral.io/health to verify server status. If issues arise, consult the MCP tooling and LangSmith platform to resolve connectivity or content access problems.

Prominent tools you can use with the server

The server exposes a rich set of prompt-related tools, including listing prompts, getting details, retrieving statistics, searching with filters, liking prompts (auth required), version history access, fetching prompt content, comparing prompts, validating prompts, and obtaining intelligent prompt completions.

Available tools

list_prompts

List public prompts with optional filtering, including pagination and basic metadata.

get_prompt

Retrieve detailed information about a specific prompt, including metadata.

get_prompt_statistics

Fetch library-wide statistics and trends for prompts and usage.

search_prompts

Perform advanced prompt searches with multiple filters such as query, owner, tags, and sort options.

like_prompt

Like a specific prompt (requires authentication) to indicate approval or usefulness.

unlike_prompt

Remove a like from a specific prompt (requires authentication).

get_prompt_versions

Get version history and commit information for a prompt.

get_user_prompts

Retrieve prompts created by a specific user.

get_popular_prompts

Fetch currently trending or highly engaged prompts.

get_prompt_content

Fetch the actual template content and configuration for a prompt.

compare_prompts

Compare multiple prompts side by side based on chosen criteria.

validate_prompt

Assess prompt structure and quality against best practices.

get_prompt_completions

Obtain intelligent auto-completions for prompt templates.