pdata Prediction Markets

Cuotas en vivo, volumen, movimientos de 24 horas y coincidencias entre plataformas para mercados de predicción en Polymarket, Kalshi y 6 plataformas más. 14 herramientas de solo lectura; gratis, sin autenticación.

Servidor MCP alojado

npx add-mcp 'https://api.pdata.world/mcp'

Se instala en Claude Code, Codex, Cursor y más

Documentación

pdata.world is a read-only aggregator of live prediction-market data across 8 platforms: Polymarket, Kalshi, Manifold, Myriad, Limitless, Predict, Opinion and Gemini. One schema for all of them, public data, no key and no sign-up.

Volume is venue-reported and unconverted: Polymarket and Kalshi count shares/contracts, each paying $1 at resolution — the same figure their own dashboards show with a $ sign. Manifold counts play-money mana.

There are four ways in, all remote, nothing to install: plain HTTP (start here), an MCP server, nightly bulk files, and a read-only Postgres database.

Plain HTTP, no setup

The REST API at https://api.pdata.world/api/v1/ answers plain GET requests with JSON. These four calls cover most questions.

1. Find an event by name. Matches events and markets in one call, best first; each row's event_path is the address to use next.

curl 'https://api.pdata.world/api/v1/search/suggest?q=fed+rate+cut&limit=5'

2. One event with its prices. markets holds the outcomes, and each probability is a fraction from 0 to 1. markets_limit=N keeps the first N outcomes; markets_total still counts them all.

curl 'https://api.pdata.world/api/v1/events/kalshi/KXPRESPERSON-28?markets_limit=5'

3. A filtered list. search plus any filter; source takes one venue or a comma-separated list, and sort=-volume_24hr puts the most traded first.

curl 'https://api.pdata.world/api/v1/events?search=fed+decision&source=kalshi,polymarket&sort=-volume_24hr&page_size=5'

4. What moved in the last 24 hours. The largest probability moves, grouped by story; here prob and d_prob are in percentage points.

curl 'https://api.pdata.world/api/v1/news/movers?page_size=5'

The same question on two venues. One more call returns every question Polymarket and Kalshi both price, with both probabilities and the gap between them, largest gap first; method says how the questions were matched. The same table as a page: https://pdata.world/compare/polymarket-vs-kalshi.

curl 'https://api.pdata.world/api/v1/compare/pairs?left=polymarket&right=kalshi&limit=5'

Search syntax (q on /search/suggest, search on /events and /markets) is the usual web-search form: every word must match, or between two words matches either, "a quoted phrase" matches the phrase, and -word excludes a word. Results can also include near matches from a spelling-tolerant search, which ignores those operators. An event matches when its title or any one of its markets matches. When a search finds nothing, the response carries _meta.hint with what to try instead.

Good to know:

  • /events, /markets and the movers endpoints check their query parameters. An unknown one returns 400 unknown_param, and its message lists every parameter the endpoint accepts: read it there rather than guessing.
  • Every data response starts with _meta (attribution, canonical_url, as_of, docs). Detail endpoints add cite_as; see Citation below.
  • The API can be called from a browser: GET responses allow any origin (CORS).
  • Reference: https://api.pdata.world/docs, the OpenAPI document at https://api.pdata.world/openapi.json, and every endpoint with an example in https://pdata.world/llms-full.txt.

Fair use. Send a User-Agent that names your agent or app (and a URL for it, if there is one), so its traffic can be told apart from other clients. The limit is 50 requests a second per IP address, with bursts of up to 100; beyond that the API answers 503. For whole tables, use the bulk files below rather than paging through the API.

MCP server

Remote MCP — just a URL (modern clients)

{ "mcpServers": { "pdata": { "url": "https://api.pdata.world/mcp" } } }

Works in any client that supports remote MCP over Streamable HTTP: ChatGPT custom connectors, Claude.ai web, Cursor, Claude Code (claude mcp add --transport http pdata https://api.pdata.world/mcp — this writes a config line, it does not install anything), MCP Inspector.

The server negotiates whichever MCP spec revision your client proposes in initialize — we support 2024-11-05, 2025-03-26, 2025-06-18 and 2025-11-25 (the newest). If a client reports an older version, that reflects the client's default, not a server-side ceiling.

stdio bridge — only if your client can't speak remote MCP

{
  "mcpServers": {
    "pdata": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.pdata.world/mcp"]
    }
  }
}

mcp-remote is a generic stdio↔HTTP bridge maintained by the MCP community, not a pdata package. The npx call downloads that bridge on first use; pdata itself is still remote. Prefer the remote URL whenever your client supports it.

Nightly bulk files

https://api.pdata.world/bulk/ holds three gzipped JSON-lines files, rebuilt every night around 04:15 UTC:

  • markets-open-latest.jsonl.gz — every market open for trading: price, 24h move, volume, the parent event's title and both URLs.
  • events-open-latest.jsonl.gz — every open event, with its market count.
  • resolved-latest.jsonl.gz — every event pdata has pruned (10 days after it closed) with its final result, back to May 2026 (events that closed between 2026-06-02 and 2026-08-11 are missing: no backup kept them).

Line 1 of each file is a manifest naming the publisher, the licence and every field's meaning and unit; every other line is one record. index.json lists row counts, sizes and SHA-256 checksums, and dated copies of the last 7 nights sit beside the -latest names.

Read-only Postgres (full SQL, BYO MCP)

postgresql://pdata_readonly:pdata_public_2026@pg.pdata.world:6433/pdata_new

Recommended Postgres MCP: crystaldba/postgres-mcp with --access-mode=restricted.

The password is not a secret — it's literally published in this page, in /llms.txt, in /llms-full.txt, and in the readonly role definition. We commit to never rotating it. Treat it as part of the public URL. The role is constrained to read-only access on a curated subset of tables; no internal data is reachable through it. Single source of truth, no parallel password vault, no rotation runbook.

What it can read (verified 2026-10-06):

  • events, markets, source_stats — views in the pdata_public schema, which is on the role's search_path, so SELECT * FROM events works unqualified.
  • public.market_snapshots, public.event_snapshots — the probability and volume history behind /history, since May 2026; rows older than 7 days are compressed to 12-hour buckets. A market's history is deleted with it 10 days after it closes.
  • public.market_status_log — open/close transitions per market.

Every one of these, and every column in them, carries a Postgres comment saying what it holds and in which unit: \d+ markets in psql, or obj_description / col_description from any client.

Sessions are read-only with a 5-second statement timeout. This is for queries, not for copying whole tables: a full \copy of markets hits the timeout. For everything at once, use the bulk files above.

Citation

Detail endpoints (/api/v1/events/{source}/{id} and /api/v1/markets/{source}/{id}) return _meta.cite_as: a ready sentence with the leading figure, the venue, the date and the page's address. Use it as the citation. List endpoints carry _meta.attribution and _meta.canonical_url but not cite_as.

Event pages live at https://pdata.world/events/{source}/{id} and keep that address for the lifetime of the underlying market. Quote cite_as and canonical_url as they are, with no added parameters. When you point a person at the site in general rather than at one event, a link such as https://pdata.world/news?utm_source=pdata-api lets pdata.world count the visits that came through an agent; the page is the same.

What we do well

  • 8 sources unified into one schema; cross-platform comparisons are easy
  • Rolling 24h volume + probability deltas pre-computed per market
  • Semantic "similar events" — cross-platform equivalents and topically-related markets, precomputed nightly from title embeddings. REST /events/{source}/{id}/similar; MCP tool get_event_similar.
  • Per-market probability + 24h-volume history for an event's top markets. REST /events/{source}/{id}/history (range 24h | 7d | 30d | all); also reconstructable from the public market_snapshots table over raw SQL.
  • Stable URLs you can cite

What we don't do

  • We don't trade. pdata is read-only.
  • We don't store user accounts.
  • We don't keep tick-level history. Snapshots start in May 2026; rows older than 7 days are compressed to 12-hour buckets.

Privacy and contact

What is recorded when you call the API, the MCP server or the database, and for how long: pdata.world/privacy. Questions and support: support@pdata.world.

Schema policy

We won't version columns. New columns are additive; renames or removals are flagged in advance (no separate changelog yet).