> ## Documentation Index
> Fetch the complete documentation index at: https://docs.valyu.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# DeepResearch

> A research API. Send a brief and get back a cited report plus the CSV, Excel, PowerPoint, Word or PDF files you ask for, with citations that can carry the passage from the source.

DeepResearch does the research job a good analyst would. You send a brief; it plans the work, runs many searches across the web and specialist sources, reads what it finds, cross-checks it and writes a cited report. Ask for the spreadsheet or the deck and it comes back too.

It runs in the background, from under 30 seconds to 90 minutes depending on mode, so your application starts a task, keeps the ID and collects the result when it's done.

## What you get back

<CardGroup cols={2}>
  <Card title="A cited report" icon="file-lines">
    Markdown or PDF, structured the way your `report_format` asks, with numbered citations to every source.
  </Card>

  <Card title="The files you asked for" icon="file-csv">
    CSV, Excel, PowerPoint, Word or PDF, up to 10 per task, generated from the same research.
  </Card>

  <Card title="The passage behind each claim" icon="quote-right">
    Sources can carry `locations`: the exact excerpt, and the page where the document has pages.
  </Card>

  <Card title="Structured data, if you prefer" icon="brackets-curly">
    Pass a JSON Schema and get validated fields instead of prose.
  </Card>
</CardGroup>

## How a task runs

<Steps>
  <Step title="Plan">
    It reads your brief, works out the sub-questions and decides which sources to use. Your `research_strategy` steers this.
  </Step>

  <Step title="Research">
    It searches and reads across the web, filings, papers, trials, patents and data, following leads as it goes. Heavy and max split a big question across up to four researchers working in parallel.
  </Step>

  <Step title="Write">
    It writes the report from its cited notes, then builds any files you asked for from the same evidence.
  </Step>
</Steps>

## Pick a mode

| Mode | Credits | Time limit | Use it for |
| - | - | - | - |
| `instant` | <span className="vx-cr">8</span> | 2 min | A short cited answer in under 30 seconds, no files |
| `fast` | <span className="vx-cr">25</span> | 10 min | Quick lookups and large batches |
| `standard` | <span className="vx-cr">100</span> | 20 min | Most research briefs |
| `heavy` | <span className="vx-cr">500</span> | 45 min | Multi-company or multi-source investigations |
| `max` | <span className="vx-cr">2,500</span> | 90 min | The largest research budget |

See [pricing](/pricing) for tools and extra files.

## New in October 2026

* **Files with the report.** Add `deliverables` and get CSV, Excel, PowerPoint, Word or PDF files back with the cited report. [Deliverables](/guides/deepresearch#deliverables)
* **Passage-level citations.** Sources can carry `locations` with the verbatim excerpt the report cited. [Citations and passages](/guides/deepresearch#citations-and-passages)
* **Instant mode.** A sourced answer in under 30 seconds for 8 credits.
* **Parallel researchers.** Heavy and max split a question across up to four researchers.
* **Pin a run to a date.** `reference_date` pins research to what was published by a past date, for backtests. [Search configuration](/guides/deepresearch#search-configuration)

The endpoint is unchanged, and older SDKs keep working. The released SDKs don't expose `locations` or `instant` yet; read them from the [task status endpoint](/api-reference/endpoint/deepresearch-status) directly.

## See it work

Real reports with their CSV files, published as they came back. [Browse all example reports](/example-reports).

<div className="vx-wf-grid not-prose">
  <a className="vx-wf" href="https://platform.valyu.ai/share/64d862ae-1d1e-4bf7-94b1-0cc8a5f22275"><span className="vx-label">Private equity · Max</span><strong>Software take-private screen</strong><span className="vx-wf-desc">Ten US-listed software companies ranked for take-private fit, every row built from SEC filings with a dated share price.</span><span className="vx-wf-ex vx-wf-meta">122 sources · 10-row CSV</span></a>
  <a className="vx-wf" href="https://platform.valyu.ai/share/bd98033e-6a9e-4f0b-bcba-4d4a3ce779c6"><span className="vx-label">Biotech & Pharma · Max</span><strong>GLP-1 drugs in Alzheimer's and Parkinson's</strong><span className="vx-wf-desc">Randomised, observational and preclinical evidence, where the field disagrees, and the registered trials it found.</span><span className="vx-wf-ex vx-wf-meta">187 sources · 24-row CSV</span></a>
</div>

## Start

<CardGroup cols={2}>
  <Card title="Run your first research report" icon="flask" href="/guides/deepresearch-quickstart">
    A walkthrough from brief to CSV in Python, TypeScript or cURL.
  </Card>

  <Card title="Configure research" icon="sliders" href="/guides/deepresearch">
    Sources, files, tools, attachments, webhooks and billing.
  </Card>

  <Card title="Use it from Claude or ChatGPT" icon="comments" href="/get-started/chat-apps">
    Ask for a DeepResearch report from your assistant, no code.
  </Card>

  <Card title="Run many at once" icon="layer-group" href="/guides/deepresearch-batch-quickstart">
    Batches with shared settings and one cost total.
  </Card>
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.