Everything we ship

This release targets the agent side of the harness: one shared query vocabulary across every tool, and a contract that lets an agent ask for exactly the slice it needs instead of pulling a whole range and discarding most of it.

• One vocabulary across every tool — query for free text, limit for result count. topK and q are retired, so an agent that has learned one search tool has learned them all.
• Ask for a slice, not a range — limit now always applies, even alongside a date range, and the new take_from picks which end you get (latest by default, or earliest). Sampling both ends of a long window is two cheap calls instead of one oversized one.
• Truncation is visible — when your window holds more than was returned, meta.notice says so. An agent can narrow the window on purpose rather than reason over a partial answer it did not know was partial.
• Mistakes fail loudly — a stale or misspelled parameter now returns an error naming its replacement instead of being quietly dropped, so a wrong call never comes back looking like a valid one.
• Stricter time inputs — dates and timestamps must be real calendar values in ISO 8601 UTC, and polymarket_price_history defaults to 720 or 365 points by interval instead of expanding to fill a wide range.

Company News is now live across LLMQuant Data. Every item carries an AI-written title, abstract, and detailed summary, plus controlled event and topic sets — so an agent can scan the market or follow a company without reading full articles. Coverage spans official SEC disclosures, press releases, and company newsroom announcements, each with a link back to the original source.

• Call news_browse from Claude, Cursor, or any MCP client. A zero-filter call returns recent market-wide news; combine tickers, events, topics, or an inclusive date range for exact filtering.
• Filter by controlled vocabularies: events such as earnings, m_and_a, or regulatory, and topics such as artificial_intelligence or banking.
• Try it from the Dashboard Sandbox — the endpoint picker now has a News group.
• Ask the Agent Playground about recent company announcements; news_browse is part of the default tool set.
• Each valid query costs 2 credits, including empty results.

See the News Browse docs for the full parameter guide and response contract.

We added plan-based API rate limits to protect normal research traffic and made the limits visible on Pricing.

• Free Beta — 30 free endpoint requests/min and 600 charged endpoint requests/min.
• Beta Pass — more room for bursty agents: 180 free endpoint requests/min and 1,200 charged endpoint requests/min.
• Clear 429s — API responses now include retry headers when you hit a limit, so your client knows when to try again.

Agents can now read the holdings and financial profile you configured under Dashboard → Profile — so a connected MCP client can answer "how concentrated is my portfolio?" before reaching for the paid market-data tools.

• personal_holdings — lists your saved positions (symbol, asset class, quantity, market value, optional cost basis, currency, as-of date). Optional asset_class filter and limit (default 30, max 50). Also available as GET /api/personal/holdings.
• personal_profile — returns your saved risk preference, investment horizon, base currency, and notes. Also available as GET /api/personal/profile.
• Read-only and scoped to you — both read only your own account and never see anyone else's data. There is no write path: agents cannot add, edit, or delete your holdings. Both cost 0 credits.

US equities now have an intraday companion to daily prices. Ask how a stock traded today or over the last few sessions — whether it gapped, reversed, or faded into the close — without pulling overly fine minute-level data.

• New equity_intraday_prices tool and GET /api/equity/intraday endpoint return 1h regular-session OHLCV bars for a single US equity.
• Pull the most recent bars with limit (default 35, ~5 trading days), or a short window with start_date + end_date (up to 14 calendar days).
• Regular session only, closed bars only — the current, still-forming bar is never included. time is an ISO 8601 UTC timestamp.
• Try it from the Dashboard Sandbox under Equity, and enable it in the Agent Playground's Equity tools when Data MCP is on.

Automations let you schedule LLMQuant's cloud agent to run a read-only financial workflow on its own — daily, weekly, or monthly — and save every run to your Agent history to read later. Each run is a one-shot report that finishes in the cloud, so nothing depends on your browser staying open.

• Open Dashboard → Automations to create one. Start from a template — Daily Market Brief, Weekly Portfolio Review, or SEC Filing Watch — or write your own prompt with date variables like {{today}} and {{week_start}}.
• Configure the schedule preset and timezone, the model, the Data MCP tools and LLMQuant Skills the run may use, and whether it can read your Personal context (off by default).
• Manage a run from its detail page: Run now, Pause, Resume, Edit, or Delete. The latest 30 runs are listed, and each finished run opens as a read-only thread in the Agent Playground.
• Glance at your automations without leaving the Agent Playground — the session rail now shows a quick preview with inline pause/resume.

Automations are a Beta Pass feature: Beta includes one active automation, and Free accounts can preview the page and draft a config. Memory is coming next.

Prediction Markets are now easier to try from the places you already work in LLMQuant Data. You can start from a finance prediction event card, read its related markets, and then pull a probability history for the outcome you care about.

• Use polymarket_event_search for natural-language questions such as Fed rate cut odds, Bitcoin ETF approval, or company earnings.
• Use polymarket_event_browse when you want a list or exact filters such as q=ETF, tag=macro, status=active, or min_volume=10000.
• Read the selected event with polymarket_event_read before choosing a child market. Then use polymarket_market_read and polymarket_price_history for market details and probability history.
• Try the full flow from the Dashboard Sandbox. Endpoint search now helps you jump straight to the Prediction Markets group.
• Ask the Agent Playground about finance prediction events and market-implied odds; the playground can now call these tools when Data MCP is enabled.
• Track Prediction Markets credit usage as its own category in the Dashboard, instead of having it grouped under Other.

See the Prediction Markets Events docs for the full event workflow and parameter guide.

LLMQuant Data is now in public Beta. Sign up, pick a plan, and start pulling agent-ready financial data in minutes — no sales call, no waitlist.

Plans & billing

• Free Beta — start free with monthly credits.
• Beta Pass — a one-time pass that adds a larger credit balance for 30 days.
• Pricing — see exactly what credits and limits each plan includes.
• One-click checkout — buy the Beta Pass through Stripe; credits land automatically.
• Dashboard — manage your key, watch usage, and grab a Beta Pass.

What you can pull today

• Market data — historical bars for equities and crypto, plus live crypto snapshots.
• Macro indicators — curated FRED series, with as-revised values.
• SEC filings — 10-K / 10-Q / 8-K sections and Form 13F holdings.
• Research — semantic search and read over quant papers.

We published LLMQuant Skills as a public repo — reusable, agent-ready workflows that turn a single ask into a full research routine.

• Browse and run them in the Agent Playground.
• Public, with new skills showing up as we ship them.

The Agent Playground is live. Ask in plain English and watch an agent pull market data, filings, macro, and research — right in the browser, no setup.

• Runs on the same key as the Data MCP server and REST API.
• See every tool call and result as the agent works.