A Source is where records enter Tapistro. Claude can create and edit 9 types of Source for you. Each section below gives one prompt for one type. Copy it into a chat with the Tapistro connector turned on, replace the [bracketed] parts, and send it.
Before you start, connect Tapistro to Claude. See Connect Tapistro to Claude.
The base prompt
Every Source prompt follows this pattern:
Create a Tapistro source.
Name: [your name for this source]
Source type: [type from the list below]
[Type-specific criteria]
Limit: max [N] accounts (or people) per run
Before saving, show me the full source config and wait for my OK.
For an account search, also add "Run a preview of 10 accounts first." Account search is the only Source type with a preview.
Source type | Use it to |
CSV upload | Import a list you already have |
Account search | Find net-new accounts that match your criteria |
Person search (recently hired) | Find people in target roles, including new hires |
Intent (Delivr) | Find accounts or people showing intent on a topic |
AI web research | Find signals no data provider tracks |
Job postings | Find accounts from hiring signals |
Email extraction | Pull accounts from alert or newsletter emails |
LinkedIn posts | Find people posting, commenting or reacting on a topic |
Account list or person list | Reuse records already in Tapistro |
1. CSV upload
Use this when you already have a list, such as event attendees, LinkedIn connections or newsletter clicks. You upload the file right in the chat, and Claude suggests the column mapping for you.
Create a CSV source named "[your name for this source]".
Send the prompt. Claude creates the Source and shows an upload box in the chat.
Click Choose a CSV or drop it here and pick your file. Files can be up to 30 MB.
Wait for the file's status to change from Uploading to Reading Columns to Mapping Ready. Claude is matching your columns to Tapistro fields.
When it shows Mapping Ready, ask Claude to review the mapping. For example: "Mapping is ready. Show me the suggested mapping."
Claude lists the mappings it suggests. Ask for any changes in plain words, or say "Save it." Once it's saved, the status changes to Approved.
Each account needs at least a domain or a company name. Each person needs at least an email or a LinkedIn URL.
To map a column to a custom field, name the field as namespace:fieldName, for example webinar:sessionAttended or event:boothVisits. The namespace groups related fields, so match it to your naming convention.
2. Account search (net-new accounts)
Create an account search source named "[your name for this source]" using [Explorium / TheirStack / Ocean.io / ContactOut].
Countries: [US, CA].
Industries: [list].
Employee count: [min] to [max].
Annual revenue: [min] to [max].
Uses these technologies: [list].
Max [N] accounts per run.
Run a preview of 10 accounts first.
Add any of these lines to narrow the search. Each one works only with the providers listed.
Say this | Filters on | Works with |
"Uses [tool]" | Tech stack | TheirStack, Ocean.io |
"Lookalikes of [domain], precise matching" | Similar companies | Ocean.io |
"Raised funding in the last [N] days" | Funding recency | TheirStack |
"Had [business event] in the last [N] days" | Business events | Explorium |
"Located in [city or region]" | Location | TheirStack |
"Website keywords: [list]" | Keywords | Explorium |
"Direct employers only" | Company type | TheirStack |
Note: Industries must match each provider's own labels. If a search comes back empty, ask Claude to check the preview.
Search by business events (Explorium)
Explorium can find companies based on something that just happened. Events cover roughly the last 90 days.
Create an account search source named "[your name for this source]" using Explorium.
Companies that had any of these events in the last [N] days: [events].
Countries: [list]. Employee count: [min] to [max].
Look up the exact Explorium event names for each event I listed, and flag any that don't exist.
Category | Events you can ask for |
Headcount growth | Increase in the sales, marketing, engineering, customer service or operations department |
Hiring | Hiring in the sales, marketing, engineering, finance, HR, legal, operations or support department |
Funding and finance | New funding round, new investment, IPO announcement |
Expansion | New office, new product, new partnership |
Organizational changes | Mergers and acquisitions, closing an office, cost cutting, a new employee joining |
Other | Company awards, lawsuits or legal issues, outages or security breaches |
3. Person search (recently hired)
Create a person and account search source named "[your name for this source]" using [Apollo / Explorium].
People: titles [list], seniority [c_suite, vp, director], located in [countries], started their role in the last [N] days.
Companies: [industries], [min] to [max] employees, in [countries].
Max [N] people per run.
"Started their role in the last [N] days" is the line that finds new hires. Leave it out to find people in those roles whatever their start date.
4. Intent (Delivr)
Delivr has an account-level version and a person-level version.
Account intent:
Create a Delivr account intent source named "[your name for this source]".
Topics: [topic names].
Pull the full list of available intent topics, find the closest match for each topic I listed, and flag any topic that doesn't exist.
Minimum intent score: high (or medium).
At least [N] people showing intent per account.
Max [N] accounts per run.
Person intent:
Create a Delivr person intent source named "[your name for this source]".
Topics: [topic names].
Pull the full list of available intent topics, find the closest match for each topic I listed, and flag any topic that doesn't exist.
Minimum intent score: high. Intent window: last [N] days.
Seniority: cxo, vp, director, manager.
Departments: [sales, marketing, operations].
Countries: [list].
Exclude employee ranges: [zero, 1 to 10, 11 to 25, 10000+].
Exclude revenue ranges: [under 1 million, 1 billion and over].
Require a business email.
Max [N] people per run.
Note: Employee and revenue ranges must use Delivr's exact labels, as shown above.
5. AI web research (web dataset)
Use this for signals that no data provider tracks, such as government contracts or niche news.
Create a web dataset source named "[your name for this source]".
Objective: find [what kind of companies or organizations] that [what you're looking for], using [specific sites or types of sources].
Each result must pass these checks:
1. [Condition you can verify as true or false]
2. [Condition]
3. [Condition] Keep only matched results.
Write each check as one statement that is clearly true or false, such as "Contract must have been awarded within the last 90 days." Claude turns each check into a match condition.
6. Job postings (hiring signals)
Create a job search source named "[your name for this source]" using Apify.
Job titles containing: [list].
Job descriptions mentioning: [list].
Posted in the last [N] days, in [countries].
Hiring company: [min] to [max] employees, in [countries], industries [list], uses [technologies].
Max [N] jobs per account and [N] accounts per run. Include full job details.
7. Email extraction (alerts and newsletters)
Create an AI entity extraction source named "[your name for this source]".
Read emails in the Gmail folder "[folder]" from sender [address], using the [connection name] Gmail connection.
Extract accounts.
This Source needs a Gmail Connection in Tapistro. See Know about and create Connections.
8. LinkedIn posts
Create a PitchGhost LinkedIn posts source named "[your name for this source]".
Search query: [boolean query, e.g., ("revops" OR "sales ops") AND "data quality"].
Include commenters and reactors.
Look back [N] days.
9. Records already in Tapistro (account list or person list)
Use these to send accounts or people you already have through a Journey, filtered by a condition.
Create an account list source named "[your name for this source]" that includes every account in Tapistro where [condition].
Create a person list source named "[your name for this source]" that includes every person where the AI agent "[agent name]" returned "[value]".
Describe the condition in plain English and Claude writes it. Three common patterns:
All records: "every account"
Records from one Source: "every account that came from the source [source name]"
An AI Agent's output: "every person where the agent [agent name] returned [value]"