> ## Documentation Index
> Fetch the complete documentation index at: https://docs.correl8.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Research

> Qualitative studies your agent runs in natural conversation

**Research** lets you ask open-ended questions through your agent — without surveys or scripted prompts. You define a study; your agent learns the answer subtly during normal conversations and posts structured answers to Correl8.

## Research guidelines

Customize how your agent conducts studies in [Guidelines](/essentials/using/guidelines) (**Research guidelines**). Per-study **question** and **eligibility criteria** are configured when you create a study below.

## Create a study

Click **New study** and fill in:

| Field                    | Purpose                                                                                               |
| ------------------------ | ----------------------------------------------------------------------------------------------------- |
| **Question**             | What you want to learn                                                                                |
| **Eligibility criteria** | When the agent should pursue this study (e.g. user type, topic, moment in the journey)                |
| **Sample percent**       | Share of users included server-side, before eligibility is even checked (deterministic per `user_id`) |
| **Max duration**         | Optional — auto-complete after N days                                                                 |
| **Max answers**          | Optional — auto-complete after N answers                                                              |

Studies start **inactive**. Open a study and click **Start** when you are ready. **Stop** pauses an active study; **Duplicate** copies settings into a new draft.

## Study status

| Status        | Meaning                                                                                                                |
| ------------- | ---------------------------------------------------------------------------------------------------------------------- |
| **Inactive**  | Not offered to users; settings fully editable                                                                          |
| **Active**    | Sampled-in users may receive the study via `get_active_studies`; question and criteria editable, sampling limits fixed |
| **Completed** | Closed — reached a limit, stopped manually, or duration expired; read-only                                             |

## How your agent participates

When a study is **active** and a user falls in the sample, the study appears in the **`get_active_studies(user_id=…)`** tool result (or the [REST studies endpoint](/essentials/active-studies-ingestion)) — sampling, duration, and answer caps are checked server-side. **Eligibility criteria are not enforced server-side**: your agent reads the criteria in the returned `guidelines` text and decides whether to pursue the study for that user. Your agent follows the loaded **research** guidelines from **`get_correl8_guidelines`** — gathering signal naturally, then calling **`post_research_answer`** (MCP) or the [REST answer endpoint](/api-reference/introduction). One answer per user per study.

Users who already answered are excluded from future `get_active_studies` results for that study.

## Review answers

Open a study to see progress, answer count, and a table of posted answers. Each row includes the agent's summary, `user_id`, `interaction_id`, and timestamp. When your project includes **Interactions**, link through to the source conversation.