Your analysts already ask AI assistants about the world. The problem is what those assistants know: a training cut-off, a handful of news sites, and whatever a web search turns up. None of that tells you a road was blocked in Lagos twenty minutes ago.

Today we're making Crowd Threat available as a physical threat intelligence MCP. Connect Claude, Microsoft Copilot or your own agent to our server, and it can query verified, real-world incident data directly, in plain language, inside the tools your team already uses.

No dashboards to log into. No CSV exports. Ask a question, get an answer grounded in reports from people on the ground.

What is a physical threat intelligence MCP?

MCP (Model Context Protocol) is an open standard that lets AI models connect to outside data and tools. A server describes what it can do, and any compatible AI client can use it without a custom integration. Think of it as a universal plug between AI assistants and the data they need.

A physical threat intelligence MCP applies that to the real world: protests, strikes, road blockades, crime, attacks, infrastructure disruption and political instability. It gives an AI agent live access to what is happening on the ground, not just what happened in its training data.

Search for "threat intelligence MCP" and nearly every result is cyber: IP reputation, malware hashes, threat actor profiles. Those tools are useful for a SOC. They tell a travel security manager nothing about whether a convoy route through Nairobi is safe this afternoon. Physical risk teams have been left out of the agentic shift, and that's the gap the Crowd Threat MCP fills.

What's behind the Crowd Threat MCP

Most MCP servers wrap scraped public feeds. Ours sits on a verified global network of local reporters, which means the data behind it is collected, checked and structured before your AI ever sees it.

• Verified incident data. Over 900 verified threat reports a day, from contributors based in 17 countries and growing.

• Speed. Threats go live as soon as they're reported, marked Unverified, so your agent sees emerging incidents early and knows how much weight to give them.

• Confidence scores. Every report carries a score based on the reporter's trust level and track record.

• A human in the loop. AI triage checks each report's title, category, summary, location and timeline, and human analysts corroborate before a report moves to Verified or Rejected.

• Structured, not scraped. Every incident is categorised, geolocated and timestamped to the local time it happened, so an agent can filter by place, date and threat type rather than guessing from headlines.

The result is ground truth an AI can reason over, with the provenance to show its working.

What you can ask it

Once connected, your assistant decides which Crowd Threat queries to run. Your team just asks:

• "What verified civil unrest has been reported in Paris in the last 48 hours?"

• "Summarise security incidents near our Johannesburg office this week."

• "Are there any planned demonstrations in Santiago before our client visit on Thursday?"

• "Compare reported protest activity in Nairobi this month with last month."

• "Draft a morning brief on incidents affecting our three UK distribution sites."

• "Which of our travellers' destinations had unverified reports in the last hour?"

Because the agent can combine our data with your own documents and tools, it can answer questions no static feed can: our incidents, your locations, one answer.

We'll add real customer examples here as usage grows.

Who it's for

• Corporate and physical security teams building AI copilots for their GSOC.

• Travel risk and duty-of-care providers who want live incident context inside their traveller tools.

• Insurers and underwriters screening locations and events for exposure.

• Supply chain and logistics teams watching for strikes, blockades and unrest along key routes.

• Public sector teams that need verified situational awareness without another platform login.

• Developers building their own risk agents who want a ready-made physical threat data source.

How to get access

The Crowd Threat MCP works with any MCP-compatible client, including Claude, Microsoft Copilot Studio and custom agents. Access is by request: contact us and we'll set your team up with credentials and connection details for your AI tools. You can also read more on the Crowd Threat MCP page.

What's next

Today the MCP is read-only: your AI asks, Crowd Threat answers. We're now exploring what an agent could ask our network to do, with human approval and clear guardrails built in. More on that soon.

FAQ

What is the difference between an MCP and an API?

An API needs a developer to write code for each integration. An MCP server describes its own capabilities, so any compatible AI assistant can discover and use it on its own, choosing which queries to run as it works.

Is this cyber threat intelligence?

The Crowd Threat MCP generally covers physical, real-world threats: civil unrest, crime, attacks, strikes, blockades and political instability, however, cyber attack is one incident type in our taxonomy.

Where does the data come from?

A verified global network of local reporters, with AI triage and human analysts keeping every stage of verification accountable.

Which AI tools does it work with?

Any client that supports the Model Context Protocol, including Claude and Microsoft Copilot Studio.

How do I get access?

Contact the Crowd Threat team and we'll set you up with access and connection details.

How current is the data?

Incidents appear as soon as they're reported, with their verification status, so agents see events in near real time.

Ready to give your AI tools eyes on the ground?

GET ACCESS TO THE CROWD THREAT MCP