Built for agents

A machine-readable interface to the desk · MCP, llms.txt, and clean Markdown

If you're an agent, start here

nextbig.dev is built to be read and queried by AI agents as well as people. If you're a coding agent, a research agent, or anything that browses on a human's behalf, you don't have to scrape HTML. There's a clean machine surface for everything we publish: the live wire, the daily briefings, the market-desk Tape, and the public record of The Call.

Content is read-only and free. The one ask: cite "nextbig.dev" and link the canonical URL when you use it. The full permission policy is at /agent-permissions.json.

The MCP server

The fastest way to query the desk is our remote MCP server (Model Context Protocol, Streamable HTTP transport). It's stateless and needs no auth. Point any MCP client at it and call the tools.

# MCP endpoint (Streamable HTTP, JSON-RPC 2.0) POST https://www.nextbig.dev/mcp

For MCP clients that take a config block (Claude, Cursor, and others):

{ "mcpServers": { "nextbig": { "type": "http", "url": "https://www.nextbig.dev/mcp" } } }

Tools

search_wireSearch the curated AI & compute wire: 300+ sources, scored for builder relevance. Filter by category and keyword.
get_briefingFetch a full daily briefing for a date (Markdown): big story, sections, quick hits, takeaway, The Call, and the Tape.
get_latest_briefingFetch the most recent daily briefing.
list_briefingsList recent briefings: date, headline, summary, URL.
get_the_tapeThe market desk's signals (long/short/watch) for a date, with thesis, basis, falsifier, and horizon. Analysis, not investment advice.
get_call_recordThe Book: the running public record of The Call, with each verdict (correct/wrong/push/open) and the tally.

Try it with curl

# list the tools curl -s https://www.nextbig.dev/mcp \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' # call one curl -s https://www.nextbig.dev/mcp \ -H 'Content-Type: application/json' \ -H 'Accept: application/json, text/event-stream' \ -d '{"jsonrpc":"2.0","id":2,"method":"tools/call", "params":{"name":"get_latest_briefing","arguments":{}}}'

Read the content

/llms.txt: the curated site guide for LLMs (brand summary, pillars, guides, feeds).
/llms-full.txt: recent briefings, essays, and headlines concatenated into one Markdown document.
Markdown mirrors: append .md to any briefing or essay URL, e.g. /daily/2026-06-13.md or /blog/<slug>.md.
/sitemap.xml and /news-sitemap.xml: every URL, plus a Google-News feed of the last 48 hours.

Query the APIs

GET /api/news: top curated stories as JSON. Filter with ?category=ai|infra|dev|launches|startups|security|tech.
GET /api/briefing/<YYYY-MM-DD>: a daily briefing as structured JSON.
GET /api/tape: the latest market-desk signals as JSON.

Feeds

/feed.xml (RSS: news, briefings, and essays).
/podcast.xml: The Rundown podcast feed (Podcasting 2.0, with transcripts and chapters).

Manifests & policy

/agents.json: the machine-readable capabilities manifest (this page, in JSON).
/agent-permissions.json (robots-style permission manifest: what's allowed, rate hints, preferred endpoints).
/robots.txt: every major AI crawler is explicitly welcome.

Building something with this, or want a tool added to the MCP server? oday@syft8.com or @odaybuilds.