mcpField
S
ShockAtlas

ShockAtlas

Remote
shockatlas.comFinance0Updated 6d ago

About ShockAtlas

ShockAtlas publishes evidence-linked signals on economic exposure to climate and geopolitical shocks. It is intended for users examining how physical events affect economic activity, with data series, causal mechanisms, hypotheses, and independently reviewed reasoning dossiers.

Key features

  1. 01Browse signals by domain and sector
  2. 02Inspect physical, economic, and flow series
  3. 03Trace typed causal links and lags
  4. 04Review dated hypotheses and updates
  5. 05Read dossiers with counter-analysis

Use cases

  1. 01Explore economic exposure to a climate shock
  2. 02Follow a shock from physical event to economic effect
  3. 03Review evidence behind an organisation’s exposure
  4. 04Check a hypothesis against published series

MCP support

The stateless Streamable HTTP MCP server exposes tools to list signals and retrieve signals, editions, dossiers, series, organisations, offers, and exports. Full exports require a member key; newsletter signup is the only write operation.

Configuration

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

Endpointhttps://shockatlas.com/api/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

Details

From the project README

ShockAtlas

An open atlas of economic exposure to climate and geopolitical shocks. Website: https://shockatlas.com · MCP server: https://shockatlas.com/api/mcp · Not investment advice.

ShockAtlas is a project of H2CO3. This repository documents what ShockAtlas publishes, how it is produced and how to query it. The site's source code is not published here.

What one signal contains

Each signal documents one shock in one place, for example the Rhine at Kaub falling below the level at which tankers can load, or a cancelled anchovy season in Peru.

  • Three public data series: one physical (a river gauge, a rainfall total, an ENSO index), one economic (a price, a freight quote), one flow (tonnes shipped, port calls, transits).
  • The mechanism: a causal chain from the physical event to the economic consequence. Each link is typed (physical, logistical, economic, behavioural), carries a lag in days, and has a status from the data: observed, consistent, untested or contradicted.
  • The organisations on the chain, each with verbatim public evidence and a sector code (NACE or NAICS).
  • Dated hypotheses with a threshold on a published series. From November 2026, each hypothesis is reviewed when its date passes and marked confirmed, invalidated or expired; a correction is published as a dated update, and the original stays visible.
  • A reasoning dossier (constants with source quotes, constraints, alternatives, conditional trajectories) reviewed by an independent counter-analyst. Disagreements are kept, not smoothed over.

Every figure on an edition is recomputed from the published observations when the site is built.

Classification

Signals are classified in a public tree: domain › sector › chain.

DomainSectors
Transportinland waterways, maritime, rail, air, road
Energygas and LNG, oil products, nuclear, hydropower, wind, solar
Agriculturecereals, tropical crops, farm inputs
Animal feedfishmeal and fish oil, oilseed meals

A sector with no signal yet is shown as "not covered yet". A chain is added only when a real signal fills it. Signals also cross the tree through shared axes: mechanism, geography (ISO 3166), series family, publisher, organisation sector. The live tree is at https://shockatlas.com/atlas.

How it is produced

ShockAtlas is run by AI agents, with separate roles that do not see each other's reasoning:

  1. Scout: picks a shock from the news or from a threshold crossed on a tracked series.
  2. Data collector: fetches the series from their publishers and records licence, publication and retrieval dates. It writes a data reading before any analysis: if the data do not show the shock, the subject stops or is reframed.
  3. Organisation mapper: finds the organisations on the chain, with verbatim public evidence.
  4. Analyst: writes the causal chain, hypotheses and trajectories.
  5. Counter-analyst: reviews the dossier from the files alone. The review is bounded to two rounds.
  6. Editor and QA: write the edition and check it against the data and the publication rules.

Data rights are checked per series. A series whose licence does not allow redistribution keeps its provenance and its source link, but its observations are withheld for everyone.

Access

Free, no key

  • Index: https://shockatlas.com/api/atlas
  • One signal: https://shockatlas.com/api/atlas/<signal>
  • A series, JSON or CSV, last 12 observations: /api/atlas/<signal>/series/<key>, /api/atlas/<signal>/csv/<key>
  • Editions, dossiers and organisations as HTML pages, an RSS feed (/feed.xml) and llms.txt
  • The free weekly newsletter: https://shockatlas.com/newsletters (double opt-in)

MCP server (streamable HTTP, stateless), listed in the official MCP registry as com.shockatlas/atlas.

claude mcp add --transport http shockatlas https://shockatlas.com/api/mcp
ToolReturns
list_signalspublished signals, filterable by domain and sector
get_signalone signal: series, figures, hypotheses, organisations
get_editionthe edition text
get_dossierthe reasoning dossier (preview without a key)
get_seriesone series (last 12 observations without a key)
get_organisationone organisation, with its sector and public evidence
get_offerswhat is free and what is paid
get_exportthe full export of a signal (member key)
subscribe_newsletterasks the free weekly for one consenting address (double opt-in); the only write

What is for sale

ShockAtlas Premium, 49 USD per month, the only paid offer:

  • every published analysis, old and new, while the subscription is active;
  • three files per analysis, and the full series and dossiers through the API and the MCP with a member key;
  • when the subscription ends, access stops; files already downloaded remain the subscriber's.

Details and current limits: https://shockatlas.com/premium

Contact

[email protected] · [email protected]

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.