MCP-native

Your prompt library,
inside your AI tools

Promptmark speaks the Model Context Protocol. Connect Claude.ai, ChatGPT, Cursor, or VS Code and manage your entire prompt library without leaving the conversation.

Start prompting — free See all features

What is the Model Context Protocol?

A standard for AI tools

MCP is an open protocol that lets AI assistants connect to external data sources and tools. Think of it as a USB port for AI — plug in any compatible service and it just works.

Two-way communication

Your AI assistant can read, create, update, and organize prompts through natural conversation. No copy-pasting. No switching tabs. Just ask.

Secure by design

Promptmark uses OAuth 2.0 with device flow authentication. Your AI tools connect to your account with scoped permissions. Your data stays yours.

74 tools. One conversation.

Every operation you can do in the Promptmark web app, you can do through MCP. Here is what your AI assistant gets access to.

Prompts

Full CRUD for your prompt library

  • list_prompts — browse with filters
  • search_prompts — find by title or tags
  • create_prompt — write new prompts
  • get_prompt — read prompt content
  • update_prompt — edit any field
  • delete_prompt — soft delete

Collections

Organize prompts into groups

  • list_collections — see all groups
  • create_collection — make new groups
  • get_collection — view with prompts
  • update_collection — rename or edit
  • delete_collection — remove groups
  • assign_prompt_to_collection — move prompts

Templates

Render prompts with variables

  • get_prompt_schema — view variable definitions
  • render_prompt — fill in template values
  • validate_prompt_inputs — check before rendering

Version history

Track every change

  • get_prompt_versions — full edit history
  • get_prompt_version — view a specific snapshot
  • restore_prompt_version — roll back changes

Captured responses

Save and review AI outputs

  • capture_response — save a model reply
  • list_captured_responses — browse saved outputs
  • get_captured_response — read full content
  • delete_captured_response — remove saved output

Tags, remixes, and safety

Organize, fork, and scan

  • list_tags — see all tags in use
  • rename_tag — bulk rename across prompts
  • list_remixes — view forked prompts
  • acknowledge_scan — resolve safety flags

Conversations

Chat with AI models directly

  • list_conversations — list with filtering and pagination
  • get_conversation — get conversation with messages
  • create_conversation — start a new conversation
  • update_conversation — update title or linked prompt
  • delete_conversation — delete a conversation
  • send_message — send a message and get AI response
  • import_conversation — import ChatGPT or Claude exports
  • list_conversation_messages — list messages in a conversation
  • link_conversation_prompt — link a prompt as system prompt

Playbooks

Define and execute multi-step workflows

  • list_playbooks — list with search and filtering
  • get_playbook — get playbook with parsed structure
  • create_playbook — create from markdown content
  • update_playbook — update with auto-versioning
  • delete_playbook — soft delete
  • validate_playbook — parse and validate markdown
  • execute_playbook — run with inputs and model
  • get_playbook_execution — execution with step results
  • list_playbook_executions — execution history
  • get_playbook_versions — version history
  • restore_playbook_version — restore previous version

Talk to your prompt library

With MCP, managing prompts feels like a conversation. Here are real things you can say to your AI assistant.

"Create a prompt called 'Code Reviewer' that checks for security issues and code style."

"Show me all my prompts tagged 'production' in the 'DevOps' collection."

"Render my 'Email Draft' template with tone='friendly' and recipient='engineering team'."

"What changed in my 'System Prompt' since last week? Restore version 3 if it looks better."

"Move all prompts tagged 'deprecated' out of the 'Active' collection."

Works with the tools you already use

Any MCP-compatible client can connect to Promptmark. These are the ones we have tested and support today.

Claude.ai

Add Promptmark via Customize → Connectors. OAuth login happens automatically.

anthropic.com

ChatGPT

Settings → Connectors → Developer Mode → Create. Paste the URL and approve.

chatgpt.com

Claude Code

One command: claude mcp add --transport http promptmark. OAuth handles the rest.

claude.ai/code

Cursor

Settings → Tools & MCP → Add New MCP Server. Paste the URL, select streamable-http.

cursor.com

VS Code

Add to .vscode/mcp.json with type http. Works with GitHub Copilot's native MCP support.

code.visualstudio.com

Warp

Settings → MCP Servers → Add. Paste the URL in the Streamable HTTP tab.

warp.dev

OpenCode

Open-source AI coding CLI with MCP support. Add to opencode.json and OAuth handles the rest.

opencode.ai

Any MCP client

Promptmark implements the full MCP spec. Just point any client at the URL.

promptmark.ai

Connect in three steps

1

Create your account

Sign up with Civic Auth. Your personal prompt database is created instantly.

2

Configure your MCP client

Add the Promptmark MCP server URL to your AI tool. OAuth login happens automatically in your browser.

3

Start talking to your prompts

Ask your AI assistant to create, find, or render prompts. All 74 tools are available immediately.

Manage prompts from your AI assistant

Free during beta. 74 MCP tools. Connect any compatible client.

Create your free account