daily.dev MCP server
Too busy to read?
Get an AI summary
Beta feature
The daily.dev MCP server is currently in beta and we're rolling it out gradually. Tools may still change. If you can't connect yet, hang tight. It's on the way.
The daily.dev MCP server exposes the public API as tools for any agent that speaks the Model Context Protocol. Add one URL to your client and your agent can read your personalized feed, search what the community is reading, save bookmarks, tune your feed filters and more, without you writing an integration.
https://api.daily.dev/mcp
It's a remote server over streamable HTTP. There is nothing to install or run locally.
What your agent gets
Every public API operation is a tool, named after its OpenAPI operationId. Some you'll see your agent reach for:
| Tool | What it does |
|---|---|
getFeedsForyou |
Your personalized feed, the same one you see on daily.dev |
getFeedsPopular, getFeedsDiscussed |
What's trending and what people are arguing about right now |
getSearchPosts, getSearchTags, getSearchSources |
Search community-vetted articles, tags and sources |
getPostsById, getPostsByIdComments |
Full post details and the comment thread |
getBookmarks, postBookmarks |
Read and save bookmarks |
getFeedsCustom, postFeedsCustom |
Read and create custom feeds |
postFeedsFiltersTagsFollow, postFeedsFiltersSourcesBlock |
Tune what shows up in your feed |
getProfile, getProfileStack |
Your profile and tech stack |
getPlugins, getPluginsById, postPlugins |
Browse and submit marketplace plugins |
Read tools are marked read-only. Tools that change your account need the write scope, which you can decline on the consent screen.
Fewer tools, less context
The server exposes around 60 tools. If your agent only needs a slice of them, point it at a marketplace plugin or the daily.dev skill instead: a SKILL.md teaches the agent the three or four calls it needs and keeps the rest out of its context window.
Connect your client
Most clients register themselves and open the daily.dev sign-in and consent screens the first time they connect. You approve access once; you can disconnect later from Settings > API > Connected apps.
Claude Code
claude mcp add --transport http daily.dev https://api.daily.dev/mcp
Then run /mcp inside Claude Code to sign in.
Claude (web and desktop)
Go to Settings > Connectors > Add custom connector, name it daily.dev and paste https://api.daily.dev/mcp. Claude opens the sign-in flow when you first use it.
Cursor
Add the server to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):
{
"mcpServers": {
"daily.dev": {
"url": "https://api.daily.dev/mcp"
}
}
}
Or install it in one click: Add daily.dev to Cursor.
Codex
codex mcp add daily-dev --url https://api.daily.dev/mcp
Codex supports sign-in with daily.dev directly. If you see "Dynamic client registration not supported", update Codex to the latest version.
VS Code
code --add-mcp '{"name":"daily.dev","type":"http","url":"https://api.daily.dev/mcp"}'
Any other client
Add a remote (streamable HTTP) server with the URL https://api.daily.dev/mcp. If your client supports OAuth, that's all you need.
Using a personal access token instead
Some clients can't run an OAuth flow, or you may be connecting from a script or a headless environment. In that case send a personal access token as a Bearer header. The token grants both read and write access.
claude mcp add --transport http daily.dev https://api.daily.dev/mcp \
--header "Authorization: Bearer $DAILY_DEV_TOKEN"
{
"mcpServers": {
"daily.dev": {
"url": "https://api.daily.dev/mcp",
"headers": {
"Authorization": "Bearer dda_your_token_here"
}
}
}
}
Keep the same rules as for the REST API: never send the token to any host other than api.daily.dev, and never commit it.
Try it
Once connected, ask your agent things like:
- "What should I read today? Pick the five most relevant posts from my daily.dev feed and tell me why."
- "What is the daily.dev community saying about Postgres 18? Give me sources."
- "Bookmark the top three posts about Rust async from this week."
- "Which tags do I follow? Suggest two I should add based on what I've been reading."
Limits and quota
MCP tool calls run as you and use the same API quota as your personal access tokens; Plus raises the limits. The MCP request itself is free; each tool call your agent makes counts as one request. Rate-limited calls return a clear error your agent can read, with the same x-ratelimit-* and retry-after headers as the REST API.
Building on the MCP server
If you're building an app that connects to the MCP server on behalf of other people, register an OAuth app and request resource=https://api.daily.dev/mcp. You don't need one to connect your own clients.