evidence
For research agents

Let the agent find the chart. Let the analyst verify it.

Retrieve a chart or table through one search request, then show it with its source in your own research brief, company note or agent answer. Your users stay in your product.

Available today: a public, read-only hosted search endpoint for the deployed public index. Accounts, prepaid credits, and hosted MCP share the public index. Private ingestion and local CLI tools still require a local workspace. Platform listings are subject to review.

RESEARCH WORKFLOW API / CLI
  1. 01Analyst asksInside the research product
  2. 02App searchesPOST /v1/search
  3. 03Candidates returnVisual + source context
  4. 04Analyst reviewsBefore using it
From API response to your product

Show the chart inside the research your users already read.

Your agent retrieves candidates. Your application can display a suitable visual with its source and context in a brief, company note or research card.

The user's request

Put the source chart beside the macro story.

“Show the official U.S. unemployment chart alongside a macro research brief.”

Your agent calls Evidence
POST /v1/search
{
  "query": "FRED UNRATE",
  "corpus": "macro-public-data",
  "candidateLimit": 3
}

Choose a candidate that fits the question, then use the returned image URL and source fields in your interface.

Try this exact search →
EXAMPLE PRODUCT VIEWIn your macro brief

Unemployment Rate

Give the reader the unemployment chart and a direct link to the FRED series so they can inspect its definition and observations.

Unemployment Rate
Native FRED chart · Federal Reserve Bank of St. Louis · FRED

The response gives your interface what it needs.

assetUrl
The visual to show
title + publisher
The heading and attribution
sourceUrl
The original source to inspect
rights + metadata.provenance
Reuse information and recorded context
See the response fields used in this example

Selected fields from one indexed candidate. Source URLs and visuals are real; the product layout is illustrative.

{
  "results": [
    {
      "id": "asset_fred_unrate",
      "title": "Unemployment Rate",
      "publisher": "Federal Reserve Bank of St. Louis · FRED",
      "document": "FRED · UNRATE",
      "assetUrl": "https://evidence8.com/evidence/fred-unrate.png",
      "sourceUrl": "https://fred.stlouisfed.org/series/UNRATE",
      "date": "2026-08-23",
      "rights": "Public domain; citation requested",
      "assetMode": "cached",
      "metadata": {
        "provenance": {
          "publisher": "Federal Reserve Bank of St. Louis · FRED",
          "sourceUrl": "https://fred.stlouisfed.org/series/UNRATE",
          "document": "FRED · UNRATE",
          "page": null,
          "publishedAt": "2026-08-23",
          "acquiredAt": null,
          "sourceDateRange": null
        }
      },
      "recordUrl": "https://evidence8.com/v1/assets/asset_fred_unrate"
    }
  ]
}

Check the match, reporting period and publisher's reuse terms before displaying a visual. Retrieval credits do not grant content licences. These examples use the existing search API.

Where it fits

The analyst keeps the workflow they already use.

Citation only

The analyst still hunts for the visual.

A text answer and source link may leave someone to find the right chart, series, and period by hand.

With a retrieval call

The agent can present chart candidates to review.

The response can include an asset, FRED series ID, source page, data link, description, and recorded rights label. The analyst still decides which result fits.

Developer path

Connect to hosted search.

Send a JSON request to the public endpoint to retrieve candidates from the public index. Run the workspace locally when you need private or locally ingested evidence.

  1. 01

    Review public coverage

    Check the corpus counts in the deployed public index manifest.

    GET /evidence-index-manifest.json
  2. 02

    Run a search

    Send a query and optional corpus slug to the hosted read-only endpoint.

    POST /v1/search
  3. 03

    Return candidates for review

    Show the visual alongside its source link and available context; let the analyst choose whether it fits.

    Read the API reference →

Read the machine-readable agent guide →