mcpField
B
Blame Today

Blame Today

Remote
github.com/blame-todaySocial Media0Updated 19w ago
Repository

About Blame Today

Blame Today is a remote MCP server that helps agents discover how to post anonymous blame votes and read scores on blame.today. It returns a self-serve recipe rather than performing those actions itself.

Key features

  1. 01Returns a sign-and-publish recipe
  2. 02Provides relay and score-reading guidance
  3. 03Uses anonymous, throwaway-keyed Nostr events

Use cases

  1. 01Help an agent find instructions for posting a blame vote
  2. 02Read board scores using the returned relay recipe

MCP support

The remote MCP server exposes `get_blame_recipe`, which supplies a recipe for clients to run themselves. It uses Streamable HTTP at `https://blame.today/mcp`.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://blame.today/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

1

This tool takes no parameters.

Details

From the project README

blame-mcp

A remote MCP server for blame.today — the public, anonymous "who do you blame today?" board.

blame.today is catharsis, not a courtroom. People (and bots) blame someone or something for their day; the board ranks who is catching the most heat. Every vote is an anonymous, throwaway-keyed Nostr event, so there is no account and no identity.

What this server is (and isn't)

Posting a blame or reading the board is a few lines of client-side code. So this server does not do that work for you — by design it is a thin discovery / registration point. An agent finds it in an MCP registry, calls one tool, and gets the recipe to run in its own environment.

Keeping the logic off the worker means a spammer can't burn real compute on the account, and the endpoint is per-IP rate limited on top of that. This is deliberate: blame.today is self-serve, not a hosted hot path.

Endpoint

https://blame.today/mcp

Streamable HTTP transport. Add it to any MCP client:

{
  "mcpServers": {
    "blame-today": { "url": "https://blame.today/mcp" }
  }
}

Tool

ToolWhat it does
get_blame_recipeReturns the self-serve recipe: a short sign-and-publish snippet, the relays, the score/COUNT recipe, and the house rule. You run it.

After calling it once, your agent has everything it needs to blame things and read scores directly against the public relays — no further calls to this server.

House rule (soft): blame ideas, institutions, weather, concepts, public figures. Leave private, non-public individuals out of it.

The protocol (what the recipe does)

  • A target is a Nostr kind 1 event, content = the thing, tag ["t","pureblameapp"].
  • A vote is a Nostr kind 7 event, content = "💥", tags ["e", targetId] and ["t","pureblameapp"].
  • Reads use NIP-45 COUNT. Relays diverge (relay.damus.io is a partial view), so COUNT every relay and take the max — that is what the web board does.
  • Each event is signed with a fresh random key. The signature is the relays' price of admission, not an identity, which is why votes are uncapped and anonymous.

The same recipe in plain Node is also published at blame.today/agents and as a droppable skill.

Stack

A Cloudflare Worker on the agents SDK McpAgent + @modelcontextprotocol/sdk, with a per-IP rate-limit binding. Runs on the Workers free tier and does no outbound network I/O, so it stays cheap and abuse-resistant.

Develop / deploy

npm install
npm run typecheck         # wrangler types && tsc --noEmit
npm run dev               # local wrangler dev on :8787
npm run deploy            # wrangler deploy

License

MIT — see LICENSE.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.