LangChain Prompts MCP Server

Measured conclusion: Deep-tested. L2 outcome: tools_ok.

Add to your agent

Copy this into your MCP client config:

{
  "mcpServers": {
    "LangChain Prompts MCP Server": {
      "url": "https://mcp.prompts.mcpcentral.io/mcp",
      "transport": "streamable-http"
    }
  }
}

Measured tools (10)

• list_prompts

  List public prompts from LangSmith with optional filtering

  Parameters:

  • limit — number — Maximum number of prompts to return
  • owner — string — Filter by prompt owner
  • search — string — Search prompts by name or description

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• get_prompt

  Get detailed information about a specific prompt

  Parameters:

  • prompt_name (required) — string — Full name of the prompt (e.g., 'owner/prompt-name')

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• get_prompt_statistics

  Get statistics about the prompt library

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• search_prompts

  Advanced search for prompts with comprehensive filtering options

  Parameters:

  • tags — array — Filter by specific tags
  • limit — number — Maximum number of results
  • owner — string — Filter by prompt owner
  • query — string — Search query for prompt names and descriptions
  • sort_by — string — Sort criteria
  • is_public — boolean — Filter by public/private status
  • min_likes — number — Minimum number of likes
  • sort_order — string — Sort order
  • min_downloads — number — Minimum number of downloads

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• get_prompt_versions

  Get version history and commits for a specific prompt

  Parameters:

  • prompt_name (required) — string — Full name of the prompt (e.g., 'owner/prompt-name')
  • limit — number — Maximum number of versions to return

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• get_popular_prompts

  Get trending and popular prompts

  Parameters:

  • limit — number — Maximum number of prompts to return
  • category — string — Filter by category/tag
  • time_period — string — Time period for popularity ranking

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• get_prompt_content

  Get the actual prompt template content and configuration

  Parameters:

  • prompt_name (required) — string — Full name of the prompt (e.g., 'owner/prompt-name')
  • version — string — Specific version/commit hash (optional, defaults to latest)
  • include_model_config — boolean — Include model configuration if available

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• compare_prompts

  Compare multiple prompts side by side

  Parameters:

  • prompt_names (required) — array — List of prompt names to compare
  • comparison_criteria — array — Criteria for comparison

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• validate_prompt

  Validate prompt structure and quality

  Parameters:

  • prompt_name (required) — string — Full name of the prompt to validate
  • check_completeness — boolean — Check if prompt has all required elements
  • check_best_practices — boolean — Check against prompt engineering best practices

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

• get_prompt_completions

  Get intelligent auto-completions for prompt templates with context awareness

  Parameters:

  • partial_text (required) — string — Partial prompt text to complete
  • context — object
  • max_suggestions — number — Maximum number of completion suggestions

  Hints: read-only not declared · destructive not declared · idempotent not declared · open-world not declared

Measured by MCPYP probing. Untested fields are omitted, never estimated. Listing is always free; MCPYP never proxies calls or stores user tokens.