UNCLASSIFIED // OSINT-DERIVED // FOUO
CRISISBRIEF
OSINT BRIEFING TERMINAL
Pro · Integrations

Connect CrisisBrief to your AI

CrisisBrief speaks the Model Context Protocol (MCP). Connect Claude, ChatGPT, Cursor or your own code, then query the holdings, pull briefs and the CrisisBrief Index, ask the cited archive, or trigger a briefing run, all from your assistant, scoped to your own data.

FASTESTOne-click: ChatGPT & Claude apps

In the ChatGPT or Claude app, add a connector with the URL https://crisisbrief.ai/api/mcp and sign in with your CrisisBrief account. No token to copy.

See the steps ↓
DEV TOOLSPersonal token: Cursor, CLI, API

For Claude Desktop, Claude Code, Cursor, the OpenAI API or any client: generate a token, then paste it into your client config.

Get a token
BEFORE YOU STARTThree things
1

You need a Pro plan.

2

Generate a token at Settings → API & MCP. Tick Allow briefing runs only if you want the assistant to start collection runs, since those draw on your monthly quota. Copy it straight away, we only show it once.

3

Your server endpoint is https://crisisbrief.ai/api/mcp. Treat the token like a password; revoke it any time from settings.

Tools

What your assistant can do

Platform intelligence
get_cbxThe current CrisisBrief Index with sub-scores and recent trend.
list_digests · get_digestPublished daily, weekly and monthly intelligence digests.
get_entity_graphThe entity co-occurrence network from published reporting.
get_track_recordThe public, Brier-scored calibration audit of resolved judgments.
get_forecast_ledger · list_forecastsThe tamper-evident, Bitcoin-anchored forecast ledger and individual sealed/resolved calls, filterable by status and outcome.
list_investigations · get_investigationLong-form deep-dive investigations in full.
Your holdings
ask_archiveA cited answer to any question, composed strictly from your claim ledger.
search_briefs · get_briefFind and read your briefs with their evidence.
search_entities · get_entity_dossierThe persistent entity registry and full dossiers.
Take action
list_cbx_alerts · set_cbx_alert · delete_cbx_alertManage CBX threshold alerts programmatically — wire index crossings into your own agents (Analyst plan).
run_briefingTrigger a new collection + analysis run (tokens with runs enabled only).
CONNECTSet up your client

Requires Node.js 18+ (for npx). Check with node -v, or install from nodejs.org.

1
Open Settings → Developer → Edit Config. That opens claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\).
2
Add the CrisisBrief block, with your token in place of cb_live_xxxxxxxxxxxx. If an mcpServers section already exists, add crisisbrief inside it.
claude_desktop_config.json
{
  "mcpServers": {
    "crisisbrief": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://crisisbrief.ai/api/mcp",
        "--header", "Authorization: Bearer cb_live_xxxxxxxxxxxx"
      ]
    }
  }
}
3
Save the file, then fully quit and reopen Claude Desktop.
4
Open the tools menu in the message box. The CrisisBrief tools should be listed. Try: “What’s the current CrisisBrief Index?”
TRY ITFirst prompts

Show me today's CrisisBrief Index and the 30-day trend.

Search my CrisisBrief briefs on the South China Sea from the last month.

Ask the CrisisBrief archive: what are the current indicators of escalation around Taiwan?

TROUBLESHOOTINGIf something is off
No tools appearConfirm Node 18+ (node -v) and fully restart the client.
401 UnauthorizedThe token is wrong or revoked. Generate a fresh one in Settings → API & MCP.
Briefing runs not allowedYour token does not have ‘Allow briefing runs’ ticked, or you have reached your monthly run quota.
Still stuckRevoke the token, generate a new one, and re-paste the config.

Bring source-graded intelligence into your workflow

UNCLASSIFIED // OSINT-DERIVED // FOUO