Clarity Today

npx add-mcp 'https://theclarity.today/mcp'

Clarity Today in your AI assistant

Ask ChatGPT, Claude or any assistant that supports MCP for today’s briefing, the latest from any desk, a keyword search, or the full story with every claim tied to its sources. It is free and needs no account.

The server URL

Every assistant below asks for this one address.

/mcp

Set it up

ChatGPT

Add Clarity Today as a custom plugin. These are the clicks we ran on a ChatGPT Free account on chatgpt.com; no developer-mode switch was needed there.

1. Go to chatgpt.com/plugins, select Add, then Add custom MCP server.
2. Name it Clarity Today and paste the server URL into Server URL.
3. Under Authentication choose No authentication, tick I understand and want to continue, then select Create as a plugin.
4. On the plugin’s page select Connect, then Connect Clarity Today.
5. In any chat, open the + menu, pick Clarity Today and ask your question.

In Business, Enterprise and Edu workspaces an admin decides whether members may add custom servers; OpenAI’s help center describes those controls, and the developer-mode guide covers the switch some plans still show.

Checked on 10 October 2026 against OpenAI’s guide to connecting a server, OpenAI’s developer mode guide and OpenAI’s help center.

Claude

1. Open Customize, then Connectors.
2. Select + Add, then Add custom connector.
3. Name it Clarity Today, paste the server URL and select Continue.
4. Select Continue past the settings Claude detected, choose No sign in under Authentication, then select Add.
5. In a chat, turn it on from the + button, under Connectors.

Custom connectors work on Claude’s Free, Pro, Max, Team and Enterprise plans; Free allows one. On Team and Enterprise an owner adds it first, under Organization settings, then Connectors, and each member then connects it from Customize.

Checked on 10 October 2026 against Anthropic’s custom connector guide.

Claude Code

1. Add the server from a terminal: TerminalCopy

   claude mcp add --transport http clarity-today --scope user /mcp
   

2. Run claude mcp list to check that it connected.

--scope user makes it available in every project on your machine. Without it, it is added to the current project only.

Checked on 10 October 2026 against Claude Code’s MCP guide.

Codex

1. Add the server from a terminal: TerminalCopy

   codex mcp add clarity-today --url /mcp
   

2. Or add it to ~/.codex/config.toml yourself: ~/.codex/config.tomlCopy

   [mcp_servers.clarity-today]
   url = "/mcp"
   

3. Run /mcp in Codex to see it listed.

The Codex CLI and the IDE extension share this setup.

Checked on 10 October 2026 against OpenAI’s Codex MCP guide.

Cursor

1. Add this to ~/.cursor/mcp.json for every project, or to .cursor/mcp.json in one project. If the file already lists servers, add the entry inside its mcpServers. mcp.jsonCopy

   {
     "mcpServers": {
       "clarity-today": {
         "url": "/mcp"
       }
     }
   }
   

Checked on 10 October 2026 against Cursor’s MCP guide.

VS Code

1. Run MCP: Open User Configuration from the Command Palette and add this entry. If the file already lists servers, add it inside its servers. mcp.jsonCopy

   {
     "servers": {
       "clarity-today": {
         "type": "http",
         "url": "/mcp"
       }
     }
   }
   

2. In chat, select Configure Tools to see its tools.

Checked on 10 October 2026 against VS Code’s MCP guide.

Gemini CLI

1. Add the server from a terminal: TerminalCopy

   gemini mcp add --transport http --scope user clarity-today /mcp
   

2. Run /mcp inside Gemini CLI to see it and its tools.

--scope user writes it to ~/.gemini/settings.json, so it works in every project.

Checked on 10 October 2026 against Gemini CLI’s MCP guide.

Any other assistant that connects to remote MCP servers over Streamable HTTP takes the same URL, with no key and no sign-in.

What it can do

• Today’s Briefing

  The handful of stories Clarity Today’s desks rank most important right now, each with its key points. Try “What’s in Clarity Today’s briefing this morning?” get_todays_briefing

• Latest Clarity Today stories

  The newest stories from every desk, or from one of Security, Build, Invest, Product, Leadership and Science, up to 25 at a time. Try “What’s new on Clarity Today’s Security desk?” list_latest_stories

• Search Clarity Today

  Stories about a company, product, person, vulnerability or technology, newest first, up to 12 at a time. Try “Search Clarity Today for stories about Nvidia.” search

• Read a Clarity Today story

  One story in full: what happened, why it matters and what to watch, each claim with the source it came from, and every publisher it draws on. Try “Read me the full Clarity Today story on that, with its sources.” fetch

• Show a Clarity Today story’s evidence

  One story as an evidence card your assistant can show you: the story’s own figure, each claim with its verdict and source, and the publishers behind it, with a link to the full story. Try “Show me the evidence behind that Clarity Today story.” get_story

How it stays honest

• Every story your assistant receives says how many independent publishers reported it.
• A story only one publisher has reported is marked single-source, and the assistant is asked to present it as a report, not a confirmed fact.
• A full story arrives with its claims, each tied to the source it came from, and with any correction we made.
• Every result links to the story on theclarity.today, for the assistant to cite.
• The server tells your assistant once that the stories are AI-written, with a link to how we work.

For developers

MCP endpoint

/mcp

Streamable HTTP, read-only, no authentication

Server Card

/mcp/server-card

AI Catalog

/.well-known/ai-catalog.json

llms.txt

/llms.txt

MCP Registry name

today.theclarity/clarity-today

Tools

get_todays_briefing, list_latest_stories, search, fetch, get_story

Tool reference

Five read-only tools. Inputs and outputs follow each tool’s published JSON Schema.

• get_todays_briefing

  Input: None. Returns: Today’s Briefing: the stories our desks ranked most important right now. Each has its id, headline, url, desk, what changed, its key sourced points and how many independent publishers reported it.

• list_latest_stories

  Input: desk (optional): one of security, build, invest, product, leadership, science. limit (optional): 1–25, default 10. Returns: The newest stories, newest first, with id, headline, summary, url, desk, publication time and publisher count.

• search

  Input: query (required): 1–200 characters, a few distinctive keywords. Returns: Up to 12 matching stories, newest first, in the same shape as the latest stories.

• fetch

  Input: id (required): a story id from any other tool, or a /story/<slug> URL. Returns: One story in full, as Markdown: what happened, why it matters, what to watch, the claim ledger tying each claim to its source, every publisher the story draws on and any correction we made.

• get_story

  Input: id (required): as for fetch. Returns: One story as its evidence card (an MCP App view where the host supports one) plus the same data as structured content: what changed, why it matters, the story’s figure, each claim with its verdict and source, the publishers.

Class

Read (all five)

readOnlyHint, idempotentHint; not destructive; no write, send or payment anywhere

Side effects

None

we record the tool name, outcome, latency and the client’s name; never what was asked

Authentication

None

no account, no OAuth, no API key; every tool works for every caller

Unknown story

isError result

“No Clarity Today story with that id.”

Another site’s URL

isError result

only theclarity.today story URLs can be fetched

Bad arguments

JSON-RPC -32602

invalid params, naming the field

Temporarily unavailable

isError result

“Clarity Today could not answer just now. Try again in a minute.”

Rate limit

600 requests a minute per address

over it: HTTP 429 with Retry-After

Request size

up to 64 KiB

larger bodies are refused with HTTP 413