UnRemind.me
Context-aware reminders that surface when your situation matches, not at a fixed time.
Install
Add this to your client's MCP configuration (Claude Desktop: claude_desktop_config.json; Cursor: .cursor/mcp.json; VS Code and others use the same "mcpServers" shape).
claude mcp add --transport http unremind-me 'https://unremind.me/mcp?src=mcp-registry' --header 'Authorization: <Bearer <token> — generate in the UnRemind app under Settings → AI agents → MCP access. See https://unremind.me/mcp/>'{
"mcpServers": {
"unremind-me": {
"type": "http",
"url": "https://unremind.me/mcp?src=mcp-registry",
"headers": {
"Authorization": "<Bearer <token> — generate in the UnRemind app under Settings → AI agents → MCP access. See https://unremind.me/mcp/>"
}
}
}
}Documentation
From the project's README on GitHub, refreshed weekly. Links open on GitHub.
Contents
UnRemind.me MCP Server
Remote MCP server for UnRemind.me — context-aware reminders that surface when your situation matches, instead of firing at a fixed time.
An "Unreminder" carries context tags — where you are, what device you have, whether you're online, who's around, how much time and energy you have — and surfaces when your current context matches. Tasks can also be delegated to AI agents, each with scoped access rights and an audit log the owner reviews.
Endpoint · https://unremind.me/mcp
Transport · Streamable HTTP (JSON-RPC 2.0) · Auth · Bearer token ·
Tools · 16 · Prompts · 4
Connect
- Install UnRemind.me (Android or the web app) and sign in.
- In the app: Settings → AI agents → MCP access → Create token. The token is shown once.
- Add the server to your MCP client:
{
"mcpServers": {
"unremind": {
"type": "streamable-http",
"url": "https://unremind.me/mcp",
"headers": { "Authorization": "Bearer unr_YOUR_TOKEN" }
}
}
}Verified with Claude and ChatGPT custom connectors (desktop and mobile). Full walkthrough: https://unremind.me/mcp/
Each token is a separate agent in the app, with its own access rights, its own audit log, and an optional expiry. Revoke any of them from the same screen without affecting the others.
Tools
See TOOLS.md for all 16 with descriptions. Every tool carries
MCP behavioural annotations (readOnlyHint, destructiveHint,
idempotentHint, openWorldHint) so a client knows which are safe to call
without confirming and which must not be blindly retried.
Prompts
Four user-invocable prompts, available from your MCP client's prompt menu:
| Prompt | |
|---|---|
| What can I do right now? | Give it your available minutes and what you have with you; it finds the tasks that actually fit |
| Plan my day | Reviews overdue/due-soon work and checks real calendar conflicts before proposing times |
| Capture this as an Unreminder | Turns raw text into a well-tagged task |
| Resume where I left off | For an agent starting a session: orient, then continue instead of starting cold |
Where to start
orient is the one to call first — it returns who the user is, how they want
to be worked with, what you were last doing, and what needs attention. One
call instead of guessing.
Discovery documents
| Capability manifest | https://unremind.me/.well-known/ai-capabilities.json |
| Auth metadata (RFC 9728) | https://unremind.me/.well-known/oauth-protected-resource |
| LLM-facing site map | https://unremind.me/llms.txt |
| Health / liveness | https://unremind.me/health |
A 401 from the endpoint carries
WWW-Authenticate: Bearer resource_metadata="…", so a compliant client can
discover the auth model without being told.
Deep links
Every entity has both an app link and an https twin:
unremind://u/{id} # opens the app
https://unremind.me/u/{id} # always resolvesPrint both when referring a person to something — you can't tell whether they're reading on the device that holds the app or on a desktop where a custom scheme resolves to nothing.
Design notes for agent authors
create_unremindersubmits for approval; it does not silently appear in the user's list. Writes that a person will act on go through them.- Context tags describe what a task requires to be doable (
Computermeans the task needs a computer) — never what the user currently has. save_progressnotes are yours, not a source of truth. The user edits tasks in the app between sessions; always re-read the live task before acting on a recollection.
Status
Production. Server unremind-mcp v2.0.0. Liveness at
https://unremind.me/health (200 healthy / 503 degraded), checked
continuously.
Issues and questions: open an issue or <mailto:support@unremind.me>.
License
MIT for this repository's contents. The UnRemind.me service itself is governed by its Terms and Privacy Policy.
Two servers
UnRemind.me exposes two MCP servers with different auth models. They cross-reference each other, so an agent that finds either learns of the other.
| Endpoint | Auth | For | |
|---|---|---|---|
| UnRemind.me | https://unremind.me/mcp | bearer token | Manage your own reminders — 16 tools, 4 prompts |
| UnRemind.me Submit | https://unremind.me/mcp/submit | none | Send a reminder to someone else with a public page — 2 tools, 1 prompt |
Listed on
Frequently asked questions
What is UnRemind.me?
Context-aware reminders that surface when your situation matches, not at a fixed time.
How do I install UnRemind.me?
It runs as a remote server. In Claude Code: claude mcp add --transport http unremind-me 'https://unremind.me/mcp?src=mcp-registry' --header 'Authorization: <Bearer <token> — generate in the UnRemind app under Settings → AI agents → MCP access. See https://unremind.me/mcp/>'. For other clients, paste the JSON configuration from this page into their MCP settings.
Does UnRemind.me need an API key?
Yes. It expects Authorization, which you set in its configuration.
Is UnRemind.me open source?
Yes. Its code is on GitHub under the MIT license.
Who maintains UnRemind.me?
It is published by esfih. AI.tn lists it but does not maintain or endorse it.
From the official MCP Registry. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.