The API for your personal memory
MyMemory saves your notes, web pages, PDFs and videos, then lets you search them and ask AI questions answered from your own saves. The REST API and MCP server give your code and your AI assistants the same library.
What you can build
A second brain for your app or agent: save what matters, then find it again by meaning.
Save anything
Send text or a link. Web pages, PDFs, YouTube videos and posts are fetched, extracted and indexed for you.
Search by meaning
Hybrid keyword and semantic search that works across languages, with date filters and recency bias.
Answers with sources
POST /chat answers from your own memories and returns citations, streamed or as one JSON reply.
Spaces and tags
Group memories into private or public spaces, tag them, and filter search and chat by either.
MCP for AI assistants
Connect Claude, ChatGPT, Cursor, VS Code and other MCP clients to the same library, local or hosted.
OpenAPI 3 spec
A full spec with real response schemas, ready for client generators and AI coding tools.
Two calls to get started
Create an API key on your MyMemory dashboard, save something, then search it. Saving is asynchronous: a memory is searchable once its processingStage is ready.
Save a page
curl https://api.mymemory.dev/v1/add \
-H "Authorization: Bearer $MYMEMORY_API_KEY" \
-H "Content-Type: application/json" \
-d '{"content": "https://example.com/article", "tags": ["reading"]}'Search your memories
curl https://api.mymemory.dev/v1/search \
-H "Authorization: Bearer $MYMEMORY_API_KEY" \
-H "Content-Type: application/json" \
-d '{"query": "what did that article say about pricing?", "limit": 5}'Endpoints
All paths are under /v1. Each links to its reference page with request and response examples.
| Method | Path | What it does |
|---|---|---|
| POST | /add | Save text or a URL as a memory |
| GET | /memories | List memories, by space or tag |
| GET | /memories/{id} | Get one memory with its content |
| PATCH | /memories/{id} | Update tags, metadata or a note |
| DELETE | /memories/{id} | Delete a memory |
| POST | /search | Hybrid keyword and semantic search |
| POST | /chat | Ask a question, answered from your memories |
| GET | /spaces | List your spaces |
| POST | /spaces/create | Create a space |
| GET | /tags | List tags with counts |
MCP server for AI assistants
Give Claude, ChatGPT, Cursor, Windsurf, VS Code and other Model Context Protocol clients five tools: save_memory, search_memories, list_spaces, get_memory_by_id and delete_memory.
Hosted (nothing to install)
https://api.mymemory.dev/v1/mcpStreamable HTTP. Authenticate with the same Authorization: Bearer sm_… header as the REST API. Works as a Claude or ChatGPT connector.
Local (stdio)
{
"mcpServers": {
"mymemory": {
"command": "npx",
"args": ["-y", "@everstash/mcp"],
"env": { "EVERSTASH_API_KEY": "sm_your_api_key" }
}
}
}Or run npx @everstash/mcp setup to configure every installed client at once. MCP guide
Plans and rate limits
The API is included in every plan, including Free. Rate limits are per minute, per endpoint; RateLimit-* headers show what's left.
| Plan | Price | Requests per minute |
|---|---|---|
| Free | $0 | 60 |
| Plus | $12/month | 300 |
| Pro | $35/month | 600 |
Yearly billing saves 17%. Storage and AI allowances are on the pricing page.