# How Research Data Integrations Work | Rippit Integration Docs > Internet information, people research, and answers from ChatGPT, Claude, and other AI assistants can all be automatically collected and stored for multi-purpose analysis. Canonical: https://www.rippit.com/integrations/docs/research-data Contents Research data integration # How research data integrations work Internet information, people research, and answers from ChatGPT, Claude, and other AI assistants can all be automatically collected and stored for multi-purpose analysis. **Collected automatically**No copying and pasting **Stored in Rippit**Kept with the rest of your data **Two kinds of sources**Web and people research, and answer engines ### How a research table works 1. 1 ### Create a research table Go to Tables > Add table > Create Research Table, or describe the research in chat. Rippit drafts the brief from your request. 2. 2 ### Write the brief The brief is the instruction every search follows: what counts as relevant, which sources to trust or skip, what evidence a result needs, and what to exclude. Put time ranges in the table settings, not the brief. 3. 3 ### Pick how it searches Web search, people search, agent search, or more than one. They share the same brief, inputs, schedule and results. 4. 4 ### Run a sample, then run it live A sample runs on up to three inputs and doesn’t affect live results or schedules. Run now writes to the live results table. ### Three ways to search - **Web search** finds pages with title, link, date, source, summary and page text. Exa by default. - **People search** finds people with name, role, company, location and profile link, and can look up work email and phone. Firecrawl by default, or Clay or Exa. - **Agent search** asks ChatGPT, Claude or Gemini questions with web search on, and records the answers, citations and searches. ### One topic or one row per item - **One topic:** the brief alone drives the search. - **Per item:** Rippit runs the brief once for each value in a column or uploaded CSV, with {{value}} as the placeholder. - Up to 100 values per table. CSVs under 5 MiB with a header row. ### Schedule and history - Run daily, weekly or by hand. - For web search, set how far back the first run looks. Later runs only fetch what’s new since each input’s last run. - Web pages and people add up across runs. Agent search keeps every study and shows the latest. ### Using the results - Ask questions about the results in chat. - Add AI Tables to classify or score rows. The research table’s own columns are fixed. - Use the table as a source for a weekly Agent App. ### Cost - AI credits for the model, plus provider costs for search. - People search contact lookups use extra provider credits. - Each run shows its reported cost, with a total across live runs. ## Research data integrations Pick a tool to see how it works. [ Research data Exa Search the web and find people that match your brief, and store every result in Rippit for analysis. ](https://www.rippit.com/integrations/exa)[ Research data Firecrawl Find people who match your brief, with work emails and phone numbers if you ask, and store them in Rippit. ](https://www.rippit.com/integrations/firecrawl)[ Research data Clay Find people who match your brief through Clay, with work emails and phone numbers if you ask, and store them in Rippit. ](https://www.rippit.com/integrations/clay)[ Research data ChatGPT Collect ChatGPT’s answers automatically, and store them in Rippit for analysis. ](https://www.rippit.com/integrations/chatgpt)[ Research data Claude Collect Claude’s answers automatically, and store them in Rippit for analysis. ](https://www.rippit.com/integrations/claude)[ Research data Gemini Collect Gemini’s answers automatically, and store them in Rippit for analysis. ](https://www.rippit.com/integrations/gemini) ### Research data integrations #### Exa Research data integration[About Exa →](https://www.rippit.com/integrations/exa) ### Use Exa 1. 1 #### Create a research table Tables > Add table > Create Research Table, or describe the research in chat. 2. 2 #### Write the brief Say what counts as relevant, which sources to trust or skip, what evidence a result needs, and what to exclude. 3. 3 #### Pick web search or people search Exa is the default for web search. For people search, choose Exa in the table settings. 4. 4 #### Run a sample, then run it live A sample covers up to three inputs and doesn’t touch live results. Who sets it up No one. Exa is built into Rippit. How it signs in No Exa account or API key Cost Included in the cost Rippit reports for each run. #### What each row holds - Web search: title, link, publish date, source, summary and full page text - People search: name, role, company, domain, location, profile link and why the person matches #### Limits - People search returns up to 100 people per input by default, up to 500 - Exa doesn’t return contact details. Work email and phone are looked up separately, only if you ask - The people search provider is fixed after the table’s first live run #### Good to know - Exa matches by meaning, not exact keywords - In chat, you can ask for up to 10 public professional profiles without creating a table. Contact lookups there are off unless you ask - Parallel, Ceramic and DeepSeek are also available for web search. Changing providers marks older samples as out of date #### Good fits - Tracking news about the companies you care about - Collecting long pages to classify or summarize with AI Tables - Building people lists before you need contact details #### Firecrawl Research data integration[About Firecrawl →](https://www.rippit.com/integrations/firecrawl) ### Use Firecrawl 1. 1 #### Create a research table Tables > Add table > Create Research Table, or describe the research in chat. 2. 2 #### Choose Firecrawl Firecrawl is the default for people search, so there’s nothing to pick. 3. 3 #### Describe the people you want Roles, seniority, company size or industry, locations, and who to exclude. To search per company, split the table on a company column and add {{value}} to the brief. 4. 4 #### Choose contact details Work email, phone, both or neither. Rippit asks before your first sample. The default is neither. 5. 5 #### Run a sample, then run it live A sample covers up to three inputs and doesn’t touch live results. Who sets it up No one. Firecrawl is built into Rippit. How it signs in No Firecrawl account or API key Cost Contact lookups use Firecrawl credits on top of AI credits. #### What each row holds - Name, role, company, company domain, location and profile link - A short reason the person matches your brief - Work email and phone, only when you ask for them, each with a lookup status #### Limits - Up to 100 people per input by default, up to 500 - Counts are maximums, not guarantees - The people search provider is fixed after the table’s first live run. To switch, create a new table #### Good to know - Firecrawl is the default people search provider - People stay in the table even when a contact lookup comes back empty - Recurring runs skip people they already found, so you don’t pay to look up the same person twice #### Clay Research data integration[About Clay →](https://www.rippit.com/integrations/clay) ### Use Clay 1. 1 #### Create a research table Tables > Add table > Create Research Table, or describe the research in chat. 2. 2 #### Choose Clay Select Clay as the people provider in the table settings, or ask for it in the table’s chat. 3. 3 #### Describe the people you want Roles, seniority, company size or industry, locations, and who to exclude. To search per company, split the table on a company column and add {{value}} to the brief. 4. 4 #### Choose contact details Work email, phone, both or neither. Rippit asks before your first sample. The default is neither. 5. 5 #### Run a sample, then run it live A sample covers up to three inputs and doesn’t touch live results. Who sets it up No one. Clay is built into Rippit. How it signs in No Clay account or API key Cost Contact lookups use Clay credits on top of AI credits. Each run reports the provider credits it used. #### What each row holds - Name, role, company, company domain, location and profile link - A short reason the person matches your brief - Work email and phone, only when you ask for them, each with a lookup status (found, outdated, not found, failed, uncertain or not requested) #### Limits - Up to 100 people per input by default, up to 500 - Counts are maximums, not guarantees - The people search provider is fixed after the table’s first live run. To switch, create a new table #### Good to know - Firecrawl is the default people search provider. Try Clay if Firecrawl’s samples look thin for your audience - To compare, run one sample with each provider on the same brief #### ChatGPT Research data integration[About ChatGPT →](https://www.rippit.com/integrations/chatgpt) ### Use ChatGPT 1. 1 #### Create a research table Tables > Add table > Create Research Table, or describe the research in chat. 2. 2 #### Turn on agent search Pick one or more OpenAI models from the Search models list. You can compare up to three models, across ChatGPT, Claude and Gemini. 3. 3 #### Write the brief Describe the topic, intent and audience. 4. 4 #### Review the questions Rippit writes 30 questions per input by default, grouped by theme. Edit them before you run. 5. 5 #### Run a sample, then run it live A sample asks up to three questions on every selected model. Each question is asked twice by default to show how consistent the answers are. Who sets it up No one. ChatGPT models are built into Rippit. How it signs in No ChatGPT account or API key Cost Uses AI credits for each call. Rippit shows the total number of calls before a run. #### What each row holds - The question and its theme - The full answer and the links it cites - The searches the model ran (when OpenAI shows them) - A full trace of the call and its status. Failed calls are kept as rows #### Good to know - These are API calls with web search on, so answers can differ from what people see in the ChatGPT app - Each run is saved as its own study. Older studies stay available under Agent search - If you change the brief, you can regenerate the questions or keep them. Questions you edited are never replaced without asking #### Good fits - Checking whether AI assistants recommend your product - Checking that answers about your product are accurate - Tracking how answers change over time on a schedule #### Claude Research data integration[About Claude →](https://www.rippit.com/integrations/claude) ### Use Claude 1. 1 #### Create a research table Tables > Add table > Create Research Table, or describe the research in chat. 2. 2 #### Turn on agent search Pick one or more Claude models from the Search models list. You can compare up to three models, across ChatGPT, Claude and Gemini. 3. 3 #### Write the brief Describe the topic, intent and audience. 4. 4 #### Review the questions Rippit writes 30 questions per input by default, grouped by theme. Edit them before you run. 5. 5 #### Run a sample, then run it live A sample asks up to three questions on every selected model. Each question is asked twice by default to show how consistent the answers are. Who sets it up No one. Claude models are built into Rippit. How it signs in No Claude account or API key Cost Uses AI credits for each call. Rippit shows the total number of calls before a run. #### What each row holds - The question and its theme - The full answer and the links it cites - The searches the model ran - A full trace of the call and its status. Failed calls are kept as rows #### Good to know - These are API calls with web search on, so answers can differ from what people see in the Claude app - Each run is saved as its own study. Older studies stay available under Agent search - If you change the brief, you can regenerate the questions or keep them. Questions you edited are never replaced without asking #### Good fits - Checking how Claude describes your product and your competitors - Comparing Claude’s cited sources with ChatGPT and Gemini #### Gemini Research data integration[About Gemini →](https://www.rippit.com/integrations/gemini) ### Use Gemini 1. 1 #### Create a research table Tables > Add table > Create Research Table, or describe the research in chat. 2. 2 #### Turn on agent search Pick one or more Gemini models from the Search models list. You can compare up to three models, across ChatGPT, Claude and Gemini. 3. 3 #### Write the brief Describe the topic, intent and audience. 4. 4 #### Review the questions Rippit writes 30 questions per input by default, grouped by theme. Edit them before you run. 5. 5 #### Run a sample, then run it live A sample asks up to three questions on every selected model. Each question is asked twice by default to show how consistent the answers are. Who sets it up No one. Gemini models are built into Rippit. How it signs in No Gemini account or API key Cost Uses AI credits for each call. Rippit shows the total number of calls before a run. #### What each row holds - The question and its theme - The full answer and the links it cites - The searches the model ran, plus grounding details - A full trace of the call and its status. Failed calls are kept as rows #### Good to know - These are API calls with Google Search on, so answers can differ from what people see in the Gemini app. They aren’t recordings of AI Overviews either. Gemini’s sources come through Google redirect links, which Rippit keeps as returned - Each run is saved as its own study. Older studies stay available under Agent search - If you change the brief, you can regenerate the questions or keep them. Questions you edited are never replaced without asking #### Good fits - Checking how Google’s AI describes your product and your competitors - Comparing which sources Gemini cites with ChatGPT and Claude ## Your data and creativity is your edge. Put agents on it to maximize it. Connect your data. Launch agents. [Start free →](https://app.rippit.com/signup)[Book a demo](https://www.rippit.com/request-demo) ## Getting started from an AI agent There are two ways to connect. Either works. - MCP server: `https://mcp.rippit.com/mcp` (Streamable HTTP, OAuth). The person signs in or signs up in the browser that opens. No client ID, secret or API key is needed. - CLI: install with `curl -fsSL https://rippit.sh | sh`, then get an account one of two ways: - `rippit login`: the person signs in or signs up in a browser. - `rippit start`: creates a temporary account with no email or browser, for a person with no Rippit account. `rippit claim --wait` lets them keep it; an unclaimed account is deleted after 14 days. Then bring in data: connect a provider such as Zendesk, Intercom, Gong or Salesforce, or upload a CSV. Full steps: https://www.rippit.com/setup.md