# Chain search → import → interrogate in one agent

> The full pipeline in one conversation: find the best sources, pull them into the library, and question them — hands off.

- For: Power user
- Category: Agent & Workflow Automation
- MCP tools: deep_search, library_import_youtube_video, deep_context
- HTML version: https://veridive.com/mcp/use-cases/search-import-interrogate-pipeline

## The problem

Every deep-research session on video content has the same three movements: find the sources worth reading, capture them somewhere stable, then interrogate them until the question is answered. Done manually — even with good tools — you're the conductor: run the search, review, trigger the imports, wait, ask the questions, repeat for the follow-ups.

The conductor role is delegable. MCP agents chain tools natively: one well-shaped instruction runs the whole pipeline — search, curate, import, interrogate — and returns with the cited answer, sources archived. You state the research goal; the movements happen.

## How veridive MCP helps

The chain is three tools your assistant sequences on its own. **deep_search** finds and ranks the candidates — quality verdicts doing the curation your review pass used to. **library_import_youtube_video** captures the top sources into your library, permanent and re-queryable. **deep_context** interrogates the imported set, returning answers **cited to the chunks**.

One prompt describes the whole arc: *"research <question>: find the strongest sources, import the best ones, then answer from them with citations."* The agent runs the movements, you receive the answer plus a durable asset — the imported sources, which every follow-up question queries for the cost of the question. It's the pattern underneath half the library-based use cases, stated once and reusable for any topic.

## How to run it

1. **Connect veridive MCP** — One-time setup: add the veridive MCP server to Claude, ChatGPT, Cursor or any MCP client and generate your API key. Takes about two minutes.
2. **State the goal, not the steps** — The agent sequences search → curate → import → interrogate from one instruction.

   ```
   Research <question>: find the strongest videos on it, import the top 5 into my library tagged '<topic>', then answer the question from those sources with citations.
   ```
3. **Follow up against the same corpus** — The imports persist — every follow-up is a scoped question, not a new pipeline.

   ```
   From the '<topic>' imports: <follow-up question>? Cited.
   ```
4. **Extend when the corpus thins** — When follow-ups outgrow the sources, the agent widens: search again, import the delta, continue.

   ```
   The current sources don't cover <angle> — find and import 3 more that do, then answer.
   ```

## Problems it solves

- Deep research on video needs a human conductor between every tool step.
- Ad-hoc research leaves nothing behind — the next question repeats the whole hunt.
- Curation-by-review eats the time the reading deserved.
- Follow-up questions get answered from memory of sources instead of the sources.

## Why it's valuable

- One instruction runs the full pipeline; you conduct outcomes, not steps.
- Every research session leaves a durable, tagged corpus behind.
- Follow-ups cost one scoped question against sources already vetted and imported.
- The pattern generalizes: any topic, same instruction shape, reusable forever.

## Example prompts

```
Research <question>: find, import top 5 tagged '<topic>', answer cited.
```

```
From '<topic>' imports: <follow-up>? Cited.
```

```
Corpus lacks <angle> — import 3 more that cover it, then answer.
```

## FAQ

**How does the agent pick which sources to import?**

deep_search's ranking and quality verdicts do the triage; your instruction sets the bar ('top 5', 'factual only', 'include niche channels'). You can also ask to see the candidates before the import step — conductor mode on demand.

**What does a full pipeline run cost?**

Search + per-video imports + the context question — a deliberate spend per research goal. The follow-ups are where it pays back: each is just a scoped question against paid-for sources.

**Is this what the domain-specific use cases are built on?**

Many of them, yes — the thesis file, the watchtower, the rationale library are this pipeline with domain framing. This page is the general pattern, stated once.

## Related use cases

- [Build an investment-thesis file from expert commentary](https://veridive.com/mcp/use-cases/investment-thesis-file.md) — Collect the strongest bull and bear cases into a library, then stress-test your thesis against all of them in one chat.
- [Learn a framework from its creator's own talks](https://veridive.com/mcp/use-cases/interrogate-the-frameworks-creator.md) — Import every talk the creator gave and ask your specific questions — the design rationale docs never capture.
- [Stand up a research library your whole team queries](https://veridive.com/mcp/use-cases/company-research-library.md) — Curate once, query forever: shared imports become a common brain with citations instead of link-dump Slack threads.

---

Part of the veridive MCP use-case library: https://veridive.com/mcp/use-cases.md
Server docs for agents: https://veridive.com/mcp.md