world-intel-mcp: an MCP collection claiming 132 tools for real-time global intel for AI agents

Verified account Tom Dörr (@tom_doerr) introduced an MCP tool collection called world-intel-mcp in a post dated October 9, 2026. The post presents the project as a way to connect AI agents to real-time global intelligence, claiming 132 MCP tools across 30 or more domains. The repository is marc-shade/world-intel-mcp.

Image: conceptual illustration for this article, not a screenshot of the repository

This article covers only what the post and its captured replies actually support. The repository README, install commands, and the real inventory behind the 132-tool claim were not part of the evidence bundle, so this piece stays within the introduction's scope and does not assert the toolset's substance.

What it is and who it is for

According to the post's wording, world-intel-mcp is a collection of Model Context Protocol tools meant to let AI agents consult outside real-time information, bundled in one repository rather than shipped as a single tool.

• Useful for: developers wiring external information tools into MCP-capable AI agents, and practitioners delegating research-style work to agents.
• Confirmed entry points: the public repository page at https://github.com/marc-shade/world-intel-mcp and the original introduction at https://x.com/tom_doerr/status/2108518050769043756.
• Caution: install commands, supported clients, licenses, and version requirements are absent from the available evidence and are therefore not stated here. Check the repository README directly before adopting it.

Why it is getting attention and what the reaction confirms

The evidence bundle contains no engagement-count fields, so this article states no like, repost, view, or bookmark figures. The observable reaction is limited to replies: three verified accounts (@gkotte1, @0x_Freak_, @asmirkn) asked about the context cost of a 132-tool bundle and whether domain-scoped loading is supported, while another verified account (@HuggingModels) left a one-line "This is a goldmine, bookmark worthy." comment. Individual replies are anecdotes, not representative measurements.

The appeal is clear: a claim of real-time global information for agents, broad coverage from a single integration, and concrete figures of 132 tools and 30-plus domains. Still, the real-time behavior and coverage are claims from the post's wording, not guarantees of performance or supported scope.

The verified limits attached to the 132-tool claim

Several verified accounts raised the same concern in the replies. Putting 132 tools on one server means the agent may have to process a large tool list before doing anything, which increases context load and the chance of calling the wrong tool. The replies ask whether tools load lazily by domain or whether the full list is passed on every call.

The key point is that no author answer to that question appears in the available evidence. So this article does not assert any of the following:

• The actual composition or current list of the 132 tools
• Whether domain-scoped lazy loading is supported
• Whether the full list is passed up front and how that affects accuracy
• The sources, latency, or coverage behind the real-time claim

If you are evaluating it, confirm the tool inventory and loading behavior in the repository first, start with a small number of domains, measure call accuracy, and only then expand.

Sources

• GitHub repository: marc-shade/world-intel-mcp
• Original post by Tom Dörr (@tom_doerr, 2026-10-09): Powers AI agents with 132 MCP tools