Find pages and documents in your own words

Requests to start from

Three ways to ask. Each one opens in the workspace, read into parts you can check and change.

1. A1

   Government reports on heat pump adoption in Europe published since 2023, as PDFs. Include the publisher and the publication date.

   Open it in the workspaceCopy for your agent
2. A2

   City council minutes from Oregon that mention short-term rental rules, since 2022. Include the city and the meeting date.

   Open it in the workspaceCopy for your agent
3. A3

   Press releases from European electric bike makers announcing new models this year. Include the link and the date.

   Open it in the workspaceCopy for your agent

How it reads

Take A1. Before the quote, SpecStream turns it into parts you can check and change. Edit one and the request saves as a new saved version. Anything it can’t support stays visible instead of being quietly dropped.

Details you can ask for

• Title
• URL
• Publisher
• Author
• Published date
• Updated date
• Language
• Document type
• File format
• Page count
• Quoted passage
• Capture date

Use them as conditions, ask for them back, or both; the field picker in the workspace shows which work where. Mark a detail required and a result needs evidence for it before it counts. Optional details can come back empty without sinking the record.

Where it looks

SpecStream picks the sources that fit each request and reads them live. Every fact keeps its source and the time it was read. Where SpecStream looks · How evidence works

From your agent or your code

Ask for the same list from your agent or your code. Your agent sees the price before it runs, then gets every match with the details you asked for. It works within your spending limit. Connect your agent with MCP.

Use SpecStream to find this:

Government reports on heat pump adoption in Europe published since 2023,
as PDFs. Include the publisher and the publication date.

Show me how you read it and the price before you run anything.

specstream compile 'Government reports on heat pump adoption in Europe published since 2023, as PDFs. Include the publisher and the publication date.'

import os
from specstream_client import Client

client = Client(api_key=os.environ["SPECSTREAM_API_KEY"])
request = client.compile(
    "Government reports on heat pump adoption in Europe published "
    "since 2023, as PDFs. Include the publisher and the publication "
    "date."
)

import { Client } from "@specstream/sdk"

const client = new Client({ apiKey: process.env.SPECSTREAM_API_KEY! })
const request = await client.compile(
    "Government reports on heat pump adoption in Europe published " +
        "since 2023, as PDFs. Include the publisher and the publication " +
        "date."
)

What it costs

You pay for the results that match, at the rate the quote shows.

• Standard, $0.10 a match: up to 5 facts to check or return, with no related list.
• Expanded, $0.30 a match: up to 15 facts, plus one related list, such as a company’s jobs or a creator’s posts. 10 related records a result are included, then 10¢ for each further 10.
• Deep, $1 a match: anything beyond Expanded.
• A person’s email, +$0.10; a phone, +$0.25 on top of the rate, when the request asks for one. A likely email counts and says likely. Rows without it pay their rate alone.
• Your own list, $0.05 a row checked. A row that matches pays its rate instead.

New accounts get $10 of work free to start, no card. See pricing

Then what?

Check the same list next quarter and see which documents changed, moved, or disappeared. Or export them to CSV, Excel, JSON, or HTML, share a snapshot with an expiry date, and pick the work up later from wherever you left it.

Keep digging

• Encyclopedia entries
• Local businesses
• Companies
• People
• Creators
• Apps
• Job postings
• Papers

Or flip through the whole crate and ask for any of them the same way.