Give your agents a prompt library
Connect your curated prompt collection to Claude, Cursor, Windsurf, and 40+ AI agents via MCP and Agent Skills. Your best prompts, always available.
The agent context problem
AI agents are powerful, but they start every session with zero context about your preferences.
Agents forget your preferences
Every new session, you re-explain your coding style, review standards, and writing preferences. Your agent has no memory of what works.
System prompts are copy-paste rituals
You paste the same system prompt into every new conversation. It's tedious, error-prone, and the prompt is never the latest version.
No single source of truth
Your best prompts live in old conversations, local files, and bookmarks. Agents can't access any of them programmatically.
Your prompt library, agent-ready
Three ways to connect your prompts to AI agents.
MCP Server
Connect via the Model Context Protocol. Agents search, read, and use your prompts directly — no copy-paste required.
Claude -> search_notes("system prompt") Agent Skills
Load prompts as Skills in Claude Code, Cursor, Windsurf. Your system prompts are loaded automatically in every session.
/pnote -> your prompts, always loaded API Tokens
Generate Personal Access Tokens for programmatic access. Build custom integrations, CI/CD pipelines, or your own agent tools.
curl -H "Authorization: Bearer pn_..." ... From prompt library to agent memory
Four steps to give your agents persistent context.
Build your prompt library
Collect system prompts, review guidelines, writing styles, and reusable patterns in PromptNote.
Connect via MCP
Add the PromptNote MCP server to your AI tool. One-time setup, always connected.
Load as Agent Skills
Install the PromptNote skill. Agents automatically access your prompt library on every invocation.
Iterate and sync
Update prompts in the web app. Changes sync instantly to all connected agents. No redeployment needed.
Compatible AI tools
Works with any tool that supports MCP or Agent Skills.
Why not hardcode system prompts?
Hardcoded prompts are static. PromptNote prompts are alive.
| Feature | PromptNote | Hardcoded Prompts |
|---|---|---|
| Update without redeployment | ✓ | — |
| Version history and diffs | ✓ | — |
| Search across all prompts | ✓ | — |
| Share across agents | ✓ | — |
| PIN protection for sensitive prompts | ✓ | — |
| Visual editing UI | ✓ | — |
What agent builders say
"MCP + PromptNote means my Claude sessions always have context. It's like giving my agent a brain that persists between conversations."
"We use Agent Skills to load review guidelines into every Claude Code session. Code quality improved measurably within a week."
"The API tokens let us pull prompts in our CI pipeline. System prompts are versioned, auditable, and always up to date."
Give your agents persistent context
Connect your prompt library via MCP or Agent Skills. Free setup, instant sync.
Frequently asked questions
What is MCP?
Model Context Protocol is an open standard that lets AI tools connect to external data sources. PromptNote's MCP server gives agents access to your prompt library.
Which agents support Skills?
Claude Code, Cursor, Windsurf, and 40+ agents that support the Skills/Rules file convention. Check our docs for the full list.
Are API tokens secure?
Yes. Tokens are stored as SHA-256 hashes. You can revoke them anytime. PIN-protected prompts remain locked even with API access.
Can multiple agents share the same prompt library?
Yes. All connected agents read from the same library. Update a prompt in the web app and every agent sees the change immediately.
Is there rate limiting?
The MCP server and API have generous rate limits suitable for individual and small team use. Enterprise plans with higher limits are planned.