Timeline Pulse ◉

The machine went quiet. The stories didn't.

Timeline Pulse is an open-source MCP server and interactive atlas of timeline-shift folklore, Mandela Effect reports, collider-era myths, and symbolic reality stories — every claim labeled, every source graded.

On 2026-06-29 the LHC was switched off for Long Shutdown 3, and won't collide again until ~2030 (CERN). Communities tell stories around machines like this. This project files those stories properly: official facts on one shelf (always cited), beliefs and symbols on another (always labeled), and tools so humans and AI agents can explore both without confusing them.

Timeline Pulse treats its material as cultural, symbolic, experiential, and speculative storytelling. It does not claim that CERN, the LHC, particle physics, timelines, or Mandela Effects prove supernatural or physical timeline manipulation.

Site: https://timeline-pulse.vercel.app
Package: timeline-pulse on npm — no keys, no backend, read-only
Built end-to-end with Claude (Claude Code + Claude in the browser designed, documented, and implemented this repo), reviewed by a human

Use it as an MCP server

config.json

{
  "mcpServers": {
    "timeline-pulse": {
      "command": "npx",
      "args": ["-y", "timeline-pulse"]
    }
  }
}

The seed corpus ships inside the package: 20 reports, 17 motifs (24 co-occurrence edges), 6 Mandela catalog items, 8 story threads, 9 timeline events with 5 officially-sourced CERN anchors.

HTTP (v2 stateless)

Default is stdio. Optional Streamable HTTP — no session id, JSON responses, loopback only:

Terminal

npx -y timeline-pulse --http
# GET  http://127.0.0.1:3000/health
# POST http://127.0.0.1:3000/mcp   (sessionless)

Env: TIMELINE_PULSE_HOST, TIMELINE_PULSE_PORT, TIMELINE_PULSE_TRANSPORT=http.

MCP tools

Tool	Purpose
`timeline_agent_manifest`	The rules of the corpus: safe language, citation policy, first calls. Call this first.
`timeline_search_reports`	Search reports by text, motif, phenomenon, source type, date, stance, or evidence grade. Paginated.
`timeline_get_report`	One full report with labeled claims, anchors, related reports, safety flags, and recommended phrasing.
`timeline_mandela_catalog`	Mandela Effect items: remembered vs on-record variants, community and mainstream notes side by side.
`timeline_motif_map`	Recurring symbols and their co-occurrence graph — the constellation, as JSON.
`timeline_story_threads`	Curated arcs written three ways: belief, neutral archive, skeptical.
`timeline_events`	Official science milestones and community waves on one axis, never blended.
`timeline_compare_sources`	Belief / neutral / skeptical framings for one topic, plus officially-sourced shared facts.
`timeline_deepen_story`	Research prompts and follow-up angles — never fabricated sources.
`timeline_stats`	Corpus totals by phenomenon, motif, source type, year, grade — and what still needs sources.

Every tool is read-only (readOnlyHint: true), local-first, and returns structured JSON with uncertainty made visible.

The site

https://timeline-pulse.vercel.app is the same corpus as a public instrument:

Archive — searchable reports with evidence-grade filters and a full-record drawer.
Timeline — official CERN anchors above the line, community waves and symbolic anchors below. One axis, never one claim.
Motifs — an interactive constellation of 17 recurring symbols (machine silence, rainbow bridge, serpent time, heart collider…).
Mandela Catalog — classic memory variants with the mainstream explanation always attached.
Story Threads — eight arcs, each readable through a belief / neutral / skeptic lens switch.
For agents — MCP install, raw JSON endpoints, llms.txt, and the language rules.

Plus one honest easter egg: press M and the site quietly Mandela-affects itself. See if you notice everything that changed.

Raw data endpoints (CORS-open)

/reports.json · /mandela-items.json · /motifs.json · /events.json · /story-threads.json · /sources.json · /llms.txt · /openapi.json

The site serves the exact JSON files the npm package ships — one corpus, three doors (site, MCP, raw JSON).

Evidence grades

Grade	Meaning
`official`	Published by the institution itself — always linked
`firsthand`	A person reporting their own experience
`reported`	Shared accounts: forums, polls, threads
`interpretive`	Community meaning-making on top of events
`symbolic`	Allegory and teaching language, filed as such
`speculative`	Unverified conjecture, clearly flagged

Development

Terminal

npm install
npm test        # policy guard + build + corpus validation + site-sync check + MCP smoke test
npm run build   # compile the server to dist/
npm run sync-site  # copy data/ JSON into site/ after editing the corpus
npx serve site  # run the site locally

The repo enforces its own content policy in CI: official claims must carry official URLs, speculative claims must carry safety flags, and one symbolic source family is never named — checked by hash, so the name appears nowhere, including in the check itself (scripts/check-policy.mjs).

Contributing

Reports, corrections, and motifs are welcome — see CONTRIBUTING.md and the issue templates. The fastest way to help: find official sources for anything flagged ⚑ needs source.

Documentation map

Project brief
MCP tool spec
Data model
Site spec
Source and content policy ← the contract
Seed corpus notes
Roadmap
Design reference

License

Code is MIT. Data: original summaries and metadata authored by the project; official facts link to their institutional sources; excerpts stay short and attributed. Details in docs/SOURCE_POLICY.md.

EST. 2026 · SEED CORPUS v0.1 · NO PHYSICS WERE HARMED ◉