AI Layoffs

Source-cited register of layoffs linked to AI, plus a live 0-100 AI job-loss index. No key.

Hosted MCP Server

npx add-mcp 'https://ailayoffs.org/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Built on Founden.ai

Layoff Data: Free Database, CSV and API | AI Layoffs

AI Layoffs

The Index Companies WARN notices Events Alerts Methodology Analysis API FAQ Build your own

Data & API

Layoff data: a free database, CSV and API

Job cuts worldwide since 2023, each with its figure, what the employer said about AI and a source anyone can open. Read the live index, its components, and the full source-cited register as one JSON endpoint, download the register as CSV, or connect an AI assistant to it over MCP. No account, no API key, CORS-enabled, because open data is the point. US layoff notices are in the WARN notice database.

No API key CORS enabled MCP server 379 source-cited events Methodology v1.36

Open the live endpoint Download the register (CSV)

Endpoints

[

GET/api/index

The live index and band, the three components with their inputs, the signals, the reconstructed history, and the full source-cited event register.

JSON

](https://ailayoffs.org/api/index)[

GET/api/register.csv

The complete source-cited register of AI-attributed layoff events, one row per event, ready for a spreadsheet or a data catalog.

](https://ailayoffs.org/api/register.csv)

AI Layoffs

An independent observatory measuring AI's impact on jobs. Neutral, source-cited, and built only on free and open data.

Product

Legal

Contact

Support© 2026 AI Layoffs. All rights reserved.

CSV

[

POST/mcp

The same register and index as a remote MCP server, so AI assistants such as Claude and ChatGPT can search it and answer with the source behind each figure. Setup below.

MCP

](#mcp)

The JSON and CSV endpoints are keyless, send Access-Control-Allow-Origin: *, and are cached at the edge for an hour, so they are safe to call directly from a browser, a notebook, or a build. The MCP endpoint is keyless and CORS-enabled too, and answers each tool call from the live data.

Quickstart

request

# the live index + full register, as JSON curl https://ailayoffs.org/api/index

just the register, as a CSV file

curl -L https://ailayoffs.org/api/register.csv -o ai-layoffs-register.csv

response (trimmed)

{ "meta": { "name": "AI Layoffs", "methodologyVersion": "1.36", "asOf": "Nov 12, 2026", "bandLabel": "Elevated", "license": "..." }, "index": { "value": 59.4, "band": { "low": 48.4, "high": 70.4 }, "trend": "down", "confidence": 0.72 }, "components": ["direct","exposure","macro"], "events": [ /* 379 source-cited events */ ] }

Connect an AI assistant (MCP)

The same register and index are served as a remote Model Context Protocol server, so Claude, ChatGPT, Cursor, VS Code and other MCP clients can query them directly and answer with the source behind every figure. Read-only, no account, no key, under the same license and citation terms.

Server URL

https://ailayoffs.org/mcp

Copy

What an assistant can ask it

get_ai_layoffs_index

The live 0-100 index reading, its band and uncertainty range, the change vs last month, and a ready-to-use citation.

count_ai_job_losses

The honest count for this year: roles disclosed, roles the employer itself tied to AI, the evidence-weighted total, and the independent Challenger figure.

search_ai_layoff_events

Find layoff events by company, text, attribution, evidence tier, execution status, country, sector, role, or date, newest first.

get_company_ai_layoffs

One company's verdict (explicit, mixed or blamed), roles disclosed vs counted, and every recorded event with its source.

Add it to your assistant

Claude (web and desktop): Settings, then Connectors, then Add custom connector. Paste the server URL; no sign-in is needed. Free plans allow one custom connector.
ChatGPT (paid plans): turn on developer mode in settings, then add a new app with the server URL.
Claude Code claude mcp add --transport http ai-layoffs https://ailayoffs.org/mcp Copy
Cursor and VS Code: one click opens the app with the server filled in. Add to Cursor Add to VS Code
By hand: Cursor and most other clients read this mcp.json config. { "mcpServers": { "ai-layoffs": { "url": "https://ailayoffs.org/mcp" } } } Copy

What is in the register

Every AI-attributed layoff event is one row, and every figure traces to its source. The index keeps two facts deliberately apart: how strong the evidence is, and how directly the employer named AI. Full definitions live on the methodology page.

company, date, sector, country

Who cut roles, when, and the company's home country.

roles counted

Roles that meet the register's standard for an AI-attributed cut. They enter the worldwide totals, and the US index sum when the roles are US-based.

roles reported (not counted)

The figure reported for the cut when the register declines to count it: stated by the company, or where the reason says so, by the press. Never added to a counted total.

why not counted

The grounds for setting a reported figure aside: the company denied AI or did not name it, or the number is not a stated reduction by the company (an upper limit, a press figure, a net change, attrition, a count that includes contractors, or one that includes roles moved elsewhere). Where the roles are is never a ground: a counted figure with no US breakdown still counts worldwide.

counts in US index

Yes only when a counted figure covers US-based roles, so it enters the US index sum.

evidence tier

How strong the underlying source is, Tier 1 (a company filing or direct statement) to Tier 3.

attribution

Explicit: the company itself tied the cut to AI. Mixed: AI named alongside other factors such as cost. Blamed: a credible source named AI but the company did not, so it is never counted.

claim + source

The exact stated reason and a link to the source for every single event.

License & attribution

Derived from free and open sources; attribute to the AI Layoffs Index and the underlying sources.

In practice: you are free to read, cite, redistribute, and build on the index and the register in research, journalism, and products, including commercial ones, as long as you attribute the AI Layoffs Index and, for a specific figure, its linked source. The full terms are on the data license page. The index recomputes live, so a durable citation carries the reading, the band, an as-of date, and the methodology version. Copy the ready-made one:

Cite this readingCopy

AI Layoffs Index: 59.4 (Elevated) as of Nov 12, 2026. AI Layoffs, methodology v1.36. https://ailayoffs.org

For a permanent, independently hosted snapshot, the source-cited register is archived with a DOI in two high-authority research repositories. On Zenodo (CERN): 10.5281/zenodo.22759361, and on Harvard Dataverse: 10.7910/DVN/ZTXNYY. Cite either for a stable identifier that resolves even if a URL changes.

To work with the register in a notebook, a copy is on Kaggle: register version 1.36, the snapshot of Oct 23, 2026. A notebook there, AI Layoffs: What Did the Employer Actually Say?, loads the file and works through who named AI as the reason, whose number each figure is and why figures are set aside. The CSV on this page is always the current one.

How the register is built and what it shows is set out in a working paper, Counting AI Layoffs from the Employer's Own Words (10.5281/zenodo.23148082), computed from register version 1.31. Cite it beside the data when you use the register's classifications.

API consumers can skip all of this and read meta.citation straight from the JSON response. Machine-readable guidance for AI systems lives at /llms.txt.

Freshness

The endpoint recomputes on every read from the live signals and the current register, and reports its own asOf date in the meta block. The register is updated as new AI-attributed events are verified against their sources.

Current reading 59.4 (Elevated) · as of Nov 12, 2026 · methodology v1.36 · data vintage October 2026

Track the index

Get told when the data changes

When the reading updates or notable new events land in the register, you get the new number, what moved, and the events, each linked to its source. Roughly monthly. Unsubscribe anytime.

Email addressGet the digest

The monthly index, the change, and the most significant new events. No spam.