There are dozens of integrations available on Hatz, and countless ways to connect your data to secure AI. Use this article to learn more about what each integration can do, and how it can be leveraged to take your AI skills to the next level.
Important Terms To Know
Integration / Tool | A connection that lets Hatz call an external service (or a built-in capability) to fetch data or perform actions.
All integrations are tools, but not all tools are integrations. Hatz has several built-in tools that do not require authentication, and dozens of integrations that DO require it. |
Tool Call | A specific function the AI can invoke (e.g., FIRECRAWL_SCRAPE, execute_python_code) to do something programmatically. Like a "menu" of possible actions that AI can take when using the integration.
Just because you connected AI to your CRM, that doesn't mean it can do ANYTHING and EVERYTHING. It has to follow a pre-determined pathway and can take actions based on the tool calls at its disposal. |
Parameters | Parameters are the named input fields you pass into a tool call so the integration knows exactly what to do.
Parameters also shape the output: they can control how many results you get, what format you get back, and what sources are allowed.
Example:
So: tool call = the action, parameters = the settings/instructions for that action. |
Authentication | Credentials required to use a third-party tool (if applicable). You might have to log in and authenticate with a third party tool to allow Hatz to connect to it.
For example, if you want to use the Microsoft integration via a Hatz Workflow, you need to authenticate with your existing Microsoft account by logging in and granting Hatz certain permissions. |
Prompt vs System Prompt |
|
Built-In Tools
What are built-in tools?
Built-in tools connect your Hatz account with software that does not require a unique account, login, or authentication. In other words, they just come with your Hatz account.
They're tools like web search, Google Maps, Code Execution, and the Hatz Workshop Assistant. So just like how you don't need to log into Google to use its Maps tool, or log into the internet to search the web, you can use these tools natively in Hatz.
Firecrawl
What Is Firecrawl?
Firecrawl is the “go read the website for me” integration: it can fetch webpages, clean up the text, and (if you want) crawl across multiple pages so you can summarize, extract, or monitor information.
The primary purpose of the Firecrawl integration is to let Hatz AI search, crawl, and scrape web pages to reliably extract their content for use in your workflows.
Read more about Firecrawl
Firecrawl Tool Calls Table
Tool call | What it does (in plain English) |
| Reads one webpage you give it and returns the content in a clean format (often markdown). |
| Reads many pages starting from a base URL (like a docs site or blog) by following links, up to a limit/depth you set. |
| If a crawl takes time, this checks progress and retrieves results when ready. |
| Stops a crawl you started (useful if it’s going too broad or you used the wrong URL). |
| Discovers a list of URLs on a site (like “show me all pages that look relevant”) before you scrape/crawl them. |
| Searches the web for your query and (optionally) scrapes the top results so you immediately get readable content. |
| Pulls specific structured fields from one or more pages (example: pricing, plan names, SSO mention) instead of returning a big blob of text. |
FAQ - How would the average person use Firecrawl?
Most people use Firecrawl to:
Summarize an article or policy page
Compare vendors by scraping pricing/feature pages
Pull key details from documentation (setup steps, requirements, limits)
Create a briefing from multiple sources (crawl a docs section or scrape a list of URLs)
What kinds of questions can I ask with Firecrawl?
Examples:
“Read this page and tell me the key takeaways.”
“Extract the pricing tiers and what’s included.”
“Crawl the docs and list all pages related to SSO.”
“Summarize changes between these two versions of a policy page.”
Is Firecrawl “read-only”?
Yes. Firecrawl’s primary job is reading/extracting content from webpages.
Can it log into sites or fill out forms?
Sometimes it can click/wait/type on a page before scraping (useful for things like dismissing popups or loading more content).
But as a general rule for non-technical users:
It cannot log in to password-protected sites.
For login-required content, you typically need a different approach (or publicly accessible pages).
Can it handle pages that load content dynamically?
Often yes—Firecrawl can wait briefly for content to load, but if a site is heavily app-like (content behind lots of scripts), results can vary.
What’s the difference between “Scrape” and “Crawl”?
Scrape = “Read this one page.”
Crawl = “Start here and read a bunch of related pages.”
When should I use “Extract” instead of “Scrape”?
Use Extract when you already know what fields you want (like a mini form):
“Plan names + prices + billing period”
“Does it mention SOC 2? Where?”
“Support hours + contact methods”
What are common mistakes?
Crawling without limits (you get too much content)
Using it on pages that block bots or are behind logins
Asking for “everything” instead of defining the outcome (summary vs. list vs. table)
Prompts Using Firecrawl (3 use cases, simple → advanced)
Use case 1 (simplest): Summarize one page
Goal: You have a link; you want the key points.
Prompt: "Use Firecrawl to scrape this URL and give me: (1) a 5-bullet summary, (2) any dates mentioned, and (3) the most important quote. URL: [PASTE URL]”
Why it works: clear output format, easy to verify.
Use case 2 (mid-level): Compare several pages in a table
Goal: You’re evaluating tools/vendors and want a structured comparison.
Prompt: “Use Firecrawl to scrape these 3 URLs and build a comparison table with columns: Pricing model, Cheapest plan, Free trial (Y/N), SSO mentioned (Y/N), SOC 2 mentioned (Y/N), and direct quotes for each claim. URLs: [URL 1], [URL 2], [URL 3]”
Why it works: forces evidence (“direct quotes”), reduces hallucinations.
Use case 3 (more complex): Crawl a docs section and extract what you care about
Goal: You want to understand a product’s documentation area without reading 40 pages yourself.
Prompt: “Use Firecrawl to crawl the documentation starting at: [DOCS HOME URL]. Limit to the docs section only. Then extract: (1) all pages related to authentication/SSO, (2) the exact setup steps mentioned, and (3) any stated limitations or prerequisites. Output a numbered checklist and include links back to the source pages.”
Why it works: crawl gathers coverage; extraction organizes it into something actionable.
Quick decision guide (one-liner)
One page →
ScrapeMany pages in a section →
Crawl(and/orMap)You want specific fields →
ExtractYou’re starting from a question, not a URL →
Searchthen scrape
Tavily
What Is Tavily?
Tavily (pronounced "ta-VILL-y") is the “go find good links on the internet, fast” integration: it helps you search the web, pull back relevant results, and optionally return a direct answer and/or raw page content to ground your work.
The primary purpose of the Tavily integration is to perform web searches and return relevant results (and sometimes direct answers) that your Hatz AI chat or workshop item can use as up-to-date external information.
Read more about Tavily
Tavily Tool Calls Table
Tool call | What it does (in plain English) |
| Searches the web for your query and returns a ranked list of relevant results (and, if you choose, a direct answer, images, and/or raw page content). |
Important Parameters
What you control in the call (common options):
query: what you’re looking formax_results: how many results to returnsearch_depth: quick search vs deeper searchinclude_answer: ask Tavily to provide a direct answer (when possible)include_raw_content: return the underlying page text (useful but can be large)include_domains/exclude_domains: restrict results to (or block) certain websites
Tavily FAQ - How would the average person use Tavily?
Most people use Tavily to:
Find credible sources quickly for a report, memo, or email
Get up-to-date context (recent announcements, policy changes, product updates)
Collect links for a competitor scan or market overview
Kickstart research before you write a summary or recommendation
Think of it like: “Google, but optimized for AI workflows.”
What kinds of questions can I ask with Tavily?
Examples:
“Find 5 reputable sources explaining X.”
“What are the latest updates on Y? Give me links from the last year.”
“Find official documentation for Z and list the key setup steps.”
Does Tavily read the whole webpage like Firecrawl?
Not by default. Tavily is primarily a search + results tool.
It can sometimes return raw content (
include_raw_content=true), but it’s not the same as a dedicated crawler/scraper.If you need clean, reliable page text, you typically pair Tavily with Firecrawl (Tavily finds the pages; Firecrawl reads them precisely).
Is Tavily “read-only”?
Yes—Tavily is for finding information, not interacting with websites.
Can Tavily log in, fill forms, or access private pages?
No. Tavily searches publicly available information. It won’t log into accounts or fill forms.
How do I keep results trustworthy?
Common best practices:
Ask for official sources first (company docs, .gov, .edu, standards orgs)
Limit results to trusted domains using
include_domainsAsk for multiple sources to confirm facts
What are common mistakes?
Asking something too broad (“Tell me everything about AI”) instead of a focused question
Not specifying constraints (timeframe, geography, “official sources only,” etc.)
Treating the search snippets as “the truth” instead of clicking/validating key claims
Prompts Using Tavily (3 use cases, simple → advanced)
Use case 1 (simplest): Get a quick list of good links
Goal: You want sources fast, not a long report.
Prompt: “Use Tavily to find 5 high-quality sources that explain [TOPIC] for business readers. Return a bullet list of links with a 1-sentence note on why each is useful.”
Why it works: it sets expectations (quality + short annotations).
Use case 2 (mid-level): Research with guardrails (official + timeframe)
Goal: You want recent, reliable info without random blogs.
Prompt: “Use Tavily to research [TOPIC] and prioritize official sources (vendor docs, regulators, standards bodies). Only include sources from the last 24 months. Return: (1) top 8 links, (2) a 10-bullet summary of key points, and (3) any disagreements between sources.”
Why it works: timeframe + source quality + cross-checking.
Use case 3 (more complex): Build a “research pack” you can hand to a team
Goal: You want a compact, decision-ready bundle of sources.
Prompt: “Use Tavily to gather a research pack on [DECISION QUESTION]. Include: pros/cons, pricing signals (if public), implementation effort, and common risks. Return a table with columns: Claim, Supporting link, Source type (official/third-party/news), and Notes. Use at least 10 sources.”
Why it works: it forces structure and keeps evidence tied to links.
Quick decision guide (one-liner)
Need links and discovery fast → Tavily
Need clean full text from a specific page → Firecrawl
Need a sourced narrative answer → Perplexity (or Exa Answer)
If you share your topic and what you’re producing (email, report, vendor comparison, etc.), I can rewrite your prompt so Tavily returns exactly the kind of sources you want.
Perplexity
What is Perplexity?
Perplexity is the “answer my question using the live web, and show me sources” integration: it can search, synthesize, and produce research-style responses that are grounded in up-to-date information.
The primary purpose of the Perplexity integration is to enable web-connected Q&A and research with sourced, up-to-date information that your Hatz AI chat or workshop item can reference.
Read more about Perplexity
Perplexity Tool Calls Table
Tool call | What it does (in plain English) |
| Runs a web search and returns ranked results (titles, URLs, snippets) so you can see what sources are out there. |
| Asks Perplexity a question and gets a direct answer that’s grounded in web info (often with sources). |
| Like “ask,” but optimized for a more careful, structured, step-by-step response. |
| Produces a deeper research write-up with citations (best for briefings, memos, and comprehensive summaries). |
Perplexity FAQ - How would the average person use it?
Most people use Perplexity to:
Get a fast, sourced explanation of a topic (with links you can share)
Produce a short research brief before a meeting
Compare options (“Which approach is better and why?”) with references
Understand “what changed recently” on a subject
If Tavily is “find me links,” Perplexity is “read around and explain it to me with sources.”
Is Perplexity read-only?
Yes. It’s designed for information retrieval and synthesis, not for logging in, filling forms, or taking actions on websites.
Can it access private/internal content?
Generally, it uses publicly available web information unless you provide text/files yourself. It’s not meant for accessing private portals or anything behind your logins.
How reliable are the citations?
Citations are helpful, but treat them like:
A map of where claims came from
Not a guarantee that every detail is perfectly interpreted
Best practice for anything important (legal, pricing, compliance, technical setup):
Use Perplexity to find and summarize
Then verify by opening the source—or pair with Firecrawl to extract the exact text/quotes from the cited pages.
When should I use search vs ask vs research?
perplexity_search: You’re still scouting sources; you want a list of links/snippets.perplexity_ask: You want a quick answer with references.perplexity_reason: You want the answer, but with more careful logic and structure.perplexity_research: You want a more complete “mini-report” with citations.
Common questions people ask
“Give me the latest on X with sources from the last year.”
“Summarize what these sources say and where they disagree.”
“Create a briefing I can send to leadership with citations.”
“What are the key pros/cons of approach A vs B?”
Prompts Using Perplexity (3 use cases, simple → advanced)
Use case 1 (simplest): Quick answer with sources
Goal: You want an accurate, shareable explanation.
Prompt: “Use Perplexity to answer: What is [TOPIC]? Keep it to 8–10 bullets and include sources for key claims.”
Why it works: short, structured, and asks for sources explicitly.
Use case 2 (mid-level): Decision support (pros/cons + risks)
Goal: You’re evaluating an approach and need a balanced view.
Prompt: “Use Perplexity to compare [OPTION A] vs [OPTION B] for a business context. Output: (1) a side-by-side table, (2) top 5 risks for each, (3) who it’s best for, and (4) citations for major claims.”
Why it works: forces a clear decision format and evidence.
Use case 3 (more complex): Research brief you can forward
Goal: You need a memo-style write-up that leadership can read.
Prompt: “Use Perplexity to write a 1–2 page briefing on [TOPIC] for a professional audience. Include: background, what’s changed in the last 12–24 months, key stakeholders, practical implications, and a ‘What to do next’ section. Use citations throughout.”
Why it works: it’s explicit about timeframe, audience, and deliverable structure.
Quick decision guide (one-liner)
Need a sourced explanation / mini-report → Perplexity
Need just a list of links → Tavily (or Perplexity Search)
Need the exact text from pages (quotes, verification, extraction) → Firecrawl
Google Search
What is Google Search?
Google Search is the “cast a wide net and find the best pages” integration: it helps you discover relevant, up-to-date webpages across the open internet so you can then read, summarize, or verify them.
The primary purpose of the Google Search integration is to let your Hatz AI chat or workshop item query Google to find relevant, up-to-date webpages and information from the web.
Read more about Google Search
Google Search Tool Calls Table
Tool call | What it does (in plain English) |
| Runs a Google-style web search for your query and returns a list of results (titles/links/snippets) to help you find relevant sources. |
Google Search FAQ - How normal people use it
Most people use it to:
Find the official page for something (docs, pricing, policies, announcements)
Quickly discover multiple sources on a topic
Pull up recent news or updates
Get a list of links to review or feed into another step (like Firecrawl)
Is it “read-only”?
Yes. Google Search is for finding pages, not interacting with websites.
Does it read the full page content?
No—this integration returns search results, not the full webpage text. If you need to actually “read” the page (to summarize accurately or quote), use Firecrawl on the URLs you found.
Can it log in or access private pages?
No. It finds public pages. Anything behind a login generally won’t be accessible.
How do I get better results?
Use constraints and specificity:
Add the company/product name and the exact thing you want (“pricing,” “SSO,” “SOC 2,” “API limits”)
Prefer official sources: add
site:company.com(or docs site)If you want a recent view: include the year or “2025”, “2026”, etc.
Exclude junk: add “-reddit” or “-pinterest” if those clutter results
Common questions people ask
“Find the official documentation page for X.”
“Find sources that explain X for business readers.”
“Find the newest policy/press release about Y.”
“Find 5 competitors to Z and their pricing pages.”
Prompts Using Google Search (3 use cases, simple → advanced)
Use case 1 (simplest): Find the official page fast
Goal: You want the one authoritative link.
Prompt: “Use Google Search to find the official documentation page for [PRODUCT/FEATURE]. Return the best 3 links and label which one is most official.”
Why it works: asks for a short list and forces “officialness.”
Use case 2 (mid-level): Build a shortlist of sources for a summary
Goal: You’re writing something and need credible references.
Prompt: “Use Google Search to find 8 reputable sources on [TOPIC]. Prioritize .gov, .edu, standards bodies, and official vendor documentation. Return a list of links with 1-line descriptions.”
Why it works: sets quality standards and a deliverable format.
Use case 3 (more complex): Research set + verification workflow
Goal: You want breadth first, then you’ll read deeply.
Prompt: “Use Google Search to find the best pages on [TOPIC], including at least: (1) an official source, (2) a recent news/announcement source, and (3) a third-party explainer. Return 10 links grouped into those categories, and suggest which 3 links I should run through Firecrawl to extract the exact text.”
Why it works: uses Google for discovery, then hands off to Firecrawl for accuracy.
Quick decision guide (one-liner)
Need to find pages broadly → Google Search
Need a fast curated search experience → Tavily
Need a sourced narrative answer → Perplexity
Need the actual page text (quotes/extraction) → Firecrawl
Exa
What is Exa?
Exa is the “find the best sources (not just keyword matches) and optionally answer with citations” integration: it’s great when you want high-quality, relevant webpages quickly—especially when your query is more conceptual than exact-keyword.
The primary purpose of the Exa integration is to power high-quality web discovery and semantic search so your Hatz AI chat or workshop item can find and cite the most relevant online sources.
Read more about Exa
Exa Tool Calls Table
Tool call | What it does (in plain English) |
| Finds and ranks relevant webpages for your query using semantic search (good for “meaning-based” discovery, not just keyword matching). |
| Produces a direct answer to your question with citations to the webpages it used (and can optionally include the source text). |
Exa FAQ - how would the average person use Exa
Most people use Exa to:
Find the most relevant pages quickly (especially when Google results feel noisy)
Discover high-quality sources for a memo or brief
Get a citation-backed answer when they want both the conclusion and where it came from
When should I use Exa instead of Google Search?
Use Exa when:
Your question is about a concept (“best practices for…”, “how companies handle…”) and keywords are messy
You want fewer, higher-quality results instead of a giant list
You want discovery that “understands the intent,” not just exact terms
Use Google Search when:
You want broad coverage, lots of options, or very specific navigation queries (“login page”, “pricing page”, “contact support”)
Is Exa read-only?
Yes. Exa is for finding and summarizing information from public webpages, not interacting with sites.
Does Exa read the full page like Firecrawl?
Not in the same way.
Exa can provide answers and citations, and sometimes include source text depending on settings.
If you need clean, reliable page text for quoting or extraction, pair Exa with Firecrawl on the selected URLs.
What kinds of questions work best with EXA_ANSWER?
Questions where you want a clear response + references, like:
“What are the main differences between X and Y?”
“What is the current consensus on Z?”
“Summarize the key requirements for [standard/policy]”
Common questions people ask
“Find the best sources on [topic] and tell me what they say.”
“Give me an answer with citations I can include in a memo.”
“Find authoritative pages that explain [concept] for business readers.”
Prompts Using Exa (3 use cases, simple → advanced)
Use case 1 (simplest): Find the best pages on a topic
Goal: You want a short, high-quality reading list.
Prompt: “Use Exa Search to find the 6 best sources on [TOPIC] for a professional audience. Return links with a 1–2 sentence summary of each.”
Why it works: Exa’s strength is high-relevance discovery.
Use case 2 (mid-level): Answer with citations (ready to paste into a doc)
Goal: You want a usable answer plus references.
Prompt: “Use Exa Answer: Explain [QUESTION] in 8–10 bullets, and include citations for each major claim. Keep the tone professional and practical.”
Why it works: tight format + explicit citations request.
Use case 3 (more complex): Research + verification plan (Exa → Firecrawl)
Goal: You want a well-sourced answer and a path to verify details.
Prompt:“Use Exa Answer to draft a short briefing on [TOPIC] with citations. Then list the top 3 URLs you used that I should scrape with Firecrawl to pull exact quotes and confirm details.”
Why it works: Exa gets you the best sources and a draft; Firecrawl can later extract exact text for accuracy.
Quick decision guide (one-liner)
Need high-quality, intent-based source discovery → Exa Search
Need a citation-backed answer → Exa Answer
Need broad discovery → Google Search
Need the exact page text / quotes / structured extraction → Firecrawl
Code Execution
What is Code Execution?
Code Execution (formerly called Python Code Interpreter) is the “let the system do the math and build the deliverable for you” integration: it runs Python in a secure workspace so you can analyze FILES, create charts, and generate downloadable documents (Excel, Word, PDF, PowerPoint, PNG, ZIP, etc.)—reliably and repeatably.
Runs Python in a sandbox to perform deterministic analysis and generate/download files (charts, tables, Word/PDF, etc.).
Code Execution Tool Calls Table
Tool call | What it does (in plain English) |
| Runs Python to process FILES, compute results, create charts, and generate output files. |
| Shows what files exist in the workspace (helpful to confirm your upload/output names). |
| Creates a download link for an output file (Word, Excel, PDF, PNG, ZIP, etc.). |
Code Execution FAQ - How would the average person use it?
Most people use it to:
Analyze spreadsheets (totals, pivots, trends, cleaning messy columns)
Create charts (bar/line/scatter) from FILE data and download as PNG/PDF
Generate deliverables (Word reports, Excel models, slide decks, PDFs)
Convert file formats (CSV ↔ Excel, tables ↔ charts, multiple files → ZIP)
The key benefit: it produces correct numbers and real files you can send to someone.
When should I use Code Execution instead of “just asking the AI”?
Use Code Execution when:
You care about accuracy (calculations, counts, grouping, filtering)
You want repeatable results (same input → same output)
You need a downloadable file (docx/xlsx/pdf/pptx/png/zip)
You’re working with real data in a FILE and don’t want manual copy/paste mistakes
What can it create (file types) and why would I choose each?
PNG (image): Best for dropping charts into email/Slack/Docs quickly.
PDF: Best for “final” read-only sharing (reports, charts) where formatting shouldn’t shift.
XLSX (Excel): Best for analysis people will keep working with (tables, pivots, formulas).
CSV: Best for simple data exchange/imports (lightweight, universal).
DOCX (Word): Best for narrative reports with headings, bullets, tables.
PPTX (PowerPoint): Best for presentations with charts + talking points.
ZIP: Best when you have multiple outputs (e.g., report + chart + cleaned spreadsheet).
Is it “read-only”?
No—this one creates and edits files in a sandbox workspace.
It won’t edit your local machine directly; it produces outputs you download.
Can it open my Excel and “click around” like a human?
Not literally. But it can read spreadsheet data and do the same work programmatically (filter, pivot, chart, export), which is often faster and less error-prone.
Common questions people ask
“Can you clean this spreadsheet and standardize the columns?”
“Can you build a chart and export it as a PNG?”
“Can you turn these notes into a Word report?”
“Can you merge these two FILES and remove duplicates?”
“Can you create an executive summary slide deck from this data?”
Prompts Using Code Execution (3 use cases, simple → advanced)
Use case 1 (simplest): Create a chart from a spreadsheet
Goal: You want a clean chart you can paste into a deck/email.
Prompt: “Use Code Execution on FILE. Create a bar chart of total revenue by region (use the columns that match revenue and region). Save as
revenue_by_region.pngand give me a download link.”What you’ll get: a real image file (PNG) generated from the data.
Use case 2 (mid-level): Clean data + output a polished Excel
Goal: You want a usable spreadsheet (not a mess) for ongoing work.
Prompt: “Use Code Execution on FILE. Clean the data by: removing duplicate rows, trimming extra spaces in text fields, standardizing date formats, and flagging missing values. Export a cleaned Excel called
cleaned_data.xlsxplus a second sheet calleddata_issueslisting any rows with problems. Give me download links.”What you’ll get: a working Excel file you can hand to a teammate.
Use case 3 (more complex): Build a full “deliverable pack” (report + charts + spreadsheet)
Goal: You want something ready to send to leadership or a client.
Prompt: “Use Code Execution on FILE to produce a deliverable pack:
summary_report.docxwith an executive summary, key metrics table, and 5 key insights,two charts saved as PNGs (trend over time + breakdown by category),
metrics_table.xlsxwith the final computed tables.
Zip everything intodeliverable_pack.zipand give me the download link.”
What you’ll get: a zip containing multiple polished outputs.
Quick decision guide (one-liner)
Need correct calculations from a FILE → Code Execution
Need charts you can download → Code Execution
Need real documents (Word/PDF/PowerPoint/Excel) → Code Execution
Need to find or read web pages → use search/scrape tools (not Code Execution)
Code Execution Demo Video
Hatz Workshop Assistant
Hatz Workshop Assistant is the “build and run AI tools inside Hatz for me” integration: it helps you create, find, copy, edit, and run Workshop items (Apps, Workflows, and Agents) without you having to wire everything manually.
Its primary function is to help you find, create, edit, copy, and run Hatz Workshop Apps/Workflows/Agents via tool calls.
Hatz Workshop Assistant Tool Calls Table
Tool call | What it does (in plain English) |
| Finds existing Apps/Workflows/Agents in Workshop that match a keyword or description. |
| Opens an item to see how it’s set up (especially its required inputs and workflow steps). |
| Creates a new App, Workflow, or Agent from your description. |
| Edits an existing item at a high level (useful for Apps/Agents or broad workflow edits). |
| Makes a safe copy of an item so you can customize it without changing the original. |
| Runs an App/Workflow/Agent to produce an output (or start a workflow job). |
| Creates an empty workflow “container” with no steps yet. |
| Adds one step to a workflow (a step is usually one prompt that produces one output). |
| Updates one specific workflow step (change prompt, add inputs, adjust dependencies). |
Hatz Workshop Assistant FAQ - How would the average person use it?
Most people use it to:
Turn a repeated task into a one-click App (summaries, email drafts, meeting notes)
Build a Workflow that does a process end-to-end (research → summarize → draft deliverable)
Create an Agent that behaves like a specialized assistant (support rep, sales copilot, recruiter helper)
Reuse someone else’s item by copying it and tailoring it to their team
What’s the difference between an App, Workflow, and Agent?
App: One-step tool. You give inputs → it returns an output right away.
Workflow: Multi-step pipeline. Step 1 output feeds step 2, etc. (good for “do A then B then C”).
Agent: Chat-based assistant with a consistent role/instructions (good for back-and-forth).
Is this “read-only”?
No—this assistant can create and modify Workshop items (especially drafts).
If you’re worried about changing something important, the safest move is:
Copy first (
COPY_WORKSHOP_ITEM), then edit the copy.
What do people usually get stuck on?
Inputs: “What does this App need from me?” → use
GET_WORKSHOP_ITEMwithinputs_only=true.Workflow step wiring: “How do I pass results from step 1 to step 2?” → use
{{outputs.step_id}}.Overbuilding: Too many inputs/steps. Most good Apps need only 3–5 inputs.
What are common questions someone might ask?
“Can you build me an App that turns bullet notes into a polished email?”
“Can you make a workflow that researches a company and creates a one-page brief?”
“Can you find an existing template for job descriptions and customize it?”
“Can you update step 2 to output JSON instead of paragraphs?”
“Why is my run failing—what inputs am I missing?”
How do I keep outputs consistent?
Tell it the exact format you want (headings, bullets, table, JSON).
Keep inputs minimal and clearly named.
For workflows, define what each step must output (so downstream steps don’t guess).
Prompts Using Hatz Workshop Assistant (3 use cases, simple → advanced)
Use case 1 (simplest): Make a one-click App for a common task
Goal: A repeatable tool for your day-to-day work.
Prompt: “Create a Workshop App called ‘Client Update Email’. Inputs: client_name, update_points (bullets), tone. Output: a subject line + a 150–250 word email. Keep it concise and professional.”
What you get: an App you (or your team) can run anytime with new inputs.
Use case 2 (mid-level): Copy an existing item and tailor it to your needs
Goal: Start from something proven; customize faster.
Prompt: “Search Workshop for an App related to ‘meeting summary’. Copy the best match into my space. Then update it so the output includes: (1) a 5-bullet summary, (2) decisions, (3) action items with owners and deadlines.”
What you get: a customized version without touching the original.
Use case 3 (more complex): Build a multi-step Workflow that produces a deliverable
Goal: A repeatable pipeline that outputs a finished artifact (brief, memo, SOP, etc.).
Prompt: “Create a Workflow called ‘Vendor Evaluation Brief’. Steps:
Summarize the vendor’s website notes I paste in (key claims, pricing, security).
Turn that into a comparison-ready table (features, risks, questions to ask).
Draft a 1-page executive brief with recommendation options.
Keep inputs minimal (vendor_name, notes, evaluation_goal).”
What you get: a workflow with clear stages so you can reuse it for any vendor.
Quick decision guide (one-liner)
Need a single reusable tool → build an App
Need a pipeline with multiple steps → build a Workflow
Need a chatty specialist you can talk to → build an Agent
Hatz Workshop Assistant Demo Video
Google Maps
What is Google Maps?
Google Maps is the “answer questions about places and travel” integration: it helps you find nearby businesses, check addresses, and estimate routes, distance, and travel time—so you can plan trips, logistics, and location-based decisions with real map data.
Retrieves place and location information (e.g., search, details, addresses, distances/routes) from Google Maps for location-based tasks.
Read more about Google Maps
Google Maps Tool Calls Table
Tool call | What it does (in plain English) |
| Searches for places using a normal text query (e.g., “coffee shops near Union Square” or “Acme HQ address”). |
| Finds places near a specific location (lat/long + radius), filtered by place type (restaurants, cafes, parks, etc.). |
| Provides turn-by-turn directions between two places (driving/walking/biking/transit), including step-by-step instructions. |
| Provides a summary route (distance + duration + overview polyline), typically used when you don’t need turn-by-turn steps. |
Google Maps FAQ - How would the average person use it?
Most people use it to:
Find addresses and basic info for places
Identify nearby options (restaurants, hotels, parking, coffee)
Estimate travel time and distance for planning
Build quick “what’s close to X?” shortlists for meetings or events
Is it read-only?
Yes—this integration reads map data (places, routes, times). It doesn’t book reservations or make purchases.
What do I need to provide to get good results?
Usually one of these:
A clear place name + city (e.g., “JFK Airport”, “Times Square NYC”)
A full address
Or coordinates (lat/long), especially for Nearby Search
What’s the difference between Text Search and Nearby Search?
Text Search: “Find X” using normal wording (best when you don’t have coordinates).
Nearby Search: “Find X within Y miles of this exact point” (best for precision).
What’s the difference between Get Route and Get Direction?
Get Route: Quick planning info (distance/time), no step-by-step.
Get Direction: Step-by-step navigation instructions (better for a detailed plan).
Can it account for traffic?
Route tools can consider traffic depending on settings; travel time may vary by time of day. If timing matters, specify when you’re leaving.
Common questions people ask
“Find the top 5 coffee shops within 10 minutes of my hotel.”
“How long does it take to get from A to B by transit?”
“What’s the best driving route that avoids tolls?”
“Find a good lunch spot near the meeting location.”
Prompts Using Google Maps (3 use cases, simple → advanced)
Use case 1 (simplest): Find an address / place details
Goal: Quickly confirm where something is.
Prompt: “Use Google Maps Text Search to find the address for [PLACE NAME]. Return the formatted address and any name variations you see.”
Why it works: direct and low friction.
Use case 2 (mid-level): Find nearby options for a meeting
Goal: Make a shortlist of places near a specific location.
Prompt: “My meeting is at [ADDRESS]. Use Google Maps to find 6 nearby cafes within a 10-minute walk. Return a shortlist with name and address.”
Why it works: clear radius + category + output format.
Use case 3 (more complex): Route planning with constraints
Goal: Plan travel with real constraints (mode, stops, avoid tolls).
Prompt: “Use Google Maps Directions from [ORIGIN] to [DESTINATION] by driving, avoiding tolls. Add a stop at [WAYPOINT]. Return total distance/time and the main steps.”
Why it works: specifies mode + constraints + waypoint + expected output.
Quick decision guide (one-liner)
Need a place/address → Text Search
Need options near a point → Nearby Search
Need step-by-step navigation → Get Direction
Need just time/distance → Get Route
Google Weather
What is Google Weather?
It's the “what’s the weather right now and what’s it going to be” integration: it pulls current conditions and multi-day forecasts for a location so you can plan travel, events, and operations.
Tool Calls Table
Tool call | What it does (in plain English) |
| Gets the current weather for a location (temperature, humidity, wind, precipitation, description). |
| Gets a daily forecast for the next N days (high/low temps, precipitation chance, conditions). |
Common inputs you control:
location: “City, State/Country” or a full addressunits: Fahrenheit (F) or Celsius (C)days(forecast only): 1–10 days
FAQ - How would the average person use Google Weather?
Most people use it to:
Check today’s conditions before commuting or traveling
Plan a trip or outdoor event (rain chances, temperature range, wind)
Pick the best day for site visits, photography, delivery schedules, or field work
Create a quick weather note for a team plan (“Expect rain Tuesday; bring gear”)
Is it read-only?
Yes. It only retrieves weather information; it doesn’t take actions (no alerts, no calendar bookings).
How specific can the location be?
You can usually use:
A city (“Chicago, IL”)
A city + country (“Paris, France”)
A street address (best when there are multiple similar place names)
What’s the difference between “current” and “forecast”?
Current: what it looks like right now (useful for “Do I need a coat?”).
Forecast: what’s expected over the next few days (useful for planning).
What information do I get back?
Typically:
Temperature
Humidity
Wind speed
Precipitation (current/expected)
A short condition label (e.g., “Cloudy,” “Rain,” “Clear”)
Common questions people ask
“Is it going to rain during my meeting window?”
“Which day this week has the lowest chance of rain?”
“What’s the temperature swing between morning and afternoon?”
“Do we expect high winds that might affect travel or setup?”
Prompts Using Google Weather (3 use cases, simple → advanced)
Use case 1 (simplest): What’s it like right now?
Goal: A quick check before leaving the house.
Prompt:
“Use Google Weather to get the current weather in LOCATION in Fahrenheit. Summarize in 2 sentences and include temperature, wind, and whether it’s raining.”
Why it works: short, practical, includes the essentials.
Use case 2 (mid-level): Pick the best day for an outdoor plan
Goal: Choose a day with good conditions.
Prompt:
“Use Google Weather forecast for LOCATION for the next 7 days in Celsius. Tell me the best 2 days for an outdoor event based on lowest rain chance and mild temperatures, and explain why.”
Why it works: makes the tool output decision-friendly.
Use case 3 (more complex): Weather brief for a trip / operations plan
Goal: A concise briefing you can share with others.
Prompt:
“Use Google Weather current + 5-day forecast for LOCATION. Create a ‘trip weather brief’ with:
Today’s conditions (now)
Daily highs/lows for the next 5 days
Rain risk callouts (which days/times look wet)
A simple packing recommendation (jacket/umbrella/light layers)”
Why it works: combines current + forecast into an actionable plan.
Quick decision guide (one-liner)
Need right now →
google_weather_currentNeed planning for days ahead →
google_weather_forecast
Hacker News
What is Hacker News?
Hacker News is the “what’s trending in tech and startups right now” integration: it lets you pull top stories, search topics, and read comment threads so you can monitor trends, discover useful links, and see what informed communities are discussing.
Pulls stories, comments, and metadata from Hacker News for trend monitoring and lightweight research.
Read more about Hacker News
Hacker News Tool Calls Table
Tool call | What it does (in plain English) |
| Pulls the current front page stories (the “top” view), optionally filtering by minimum points. |
| Pulls stories from today that meet a points threshold (useful for daily trend snapshots). |
| Pulls the most recent posts (new submissions), optionally filtered by tags (e.g., Ask HN, Show HN). |
| Searches Hacker News posts by keyword (great for finding discussions on a topic). |
| Opens a specific story or comment by ID, including a limited set of comments (thread reading). |
| Retrieves information about a specific HN user (profile basics, karma, etc.). |
Hacker News FAQ - How would the average person use it?
Most people use it to:
Get a daily pulse on what’s popular in tech
Find high-signal articles that are being widely discussed
Read comments to understand pros/cons, skepticism, and real-world experience
Track topics (AI tools, security incidents, developer trends, startups)
Is it read-only?
Yes. This integration reads Hacker News data (posts, comments, users). It doesn’t post or vote.
What’s the difference between “front page” and “latest”?
Front page: What’s currently most popular (higher signal, more discussion).
Latest: What’s newest (good for spotting early trends, but noisier).
What are “Ask HN” and “Show HN”?
Ask HN: Questions posted to the community (often practical advice).
Show HN: People sharing something they built (products, demos, open source).
How should I use HN comments?
Comments are valuable for:
Real-world feedback (“we tried this, here’s what happened”)
Spotting missing context or weaknesses But remember: it’s still opinions—use comments as input, not as your only source of truth.
Common questions people ask
“What are the top 10 stories today in AI/security/dev tools?”
“Find recent discussions about [company/product].”
“Summarize the debate in this thread—what are the main viewpoints?”
“What’s the strongest criticism people raised about this announcement?”
Prompts Using Hacker News (3 use cases, simple → advanced)
Use case 1 (simplest): Daily trend snapshot
Goal: Quickly see what matters today.
Prompt: “Use Hacker News to pull today’s top posts (minimum 200 points). Give me the top 10 with a one-line takeaway for each.”
Why it works: a clean daily briefing format.
Use case 2 (mid-level): Track a topic over time
Goal: Find relevant discussions about a specific topic.
Prompt: “Search Hacker News for posts about [TOPIC] from the last few years (or most relevant). Return 8 of the most useful threads and explain why each is worth reading.”
Why it works: turns search results into an actionable reading list.
Use case 3 (more complex): Summarize a comment thread like a mini focus group
Goal: Extract the main arguments from a discussion.
Prompt: “Open this Hacker News story by ID: [ID]. Summarize the top viewpoints in the comments: (1) strongest arguments in favor, (2) strongest criticisms, (3) any real-world examples, and (4) open questions. Quote 2–3 short excerpts and include comment links when possible.”
Why it works: treats HN as “qualitative research” and asks for balanced synthesis.
Quick decision guide (one-liner)
Want what’s popular right now → Front page / Today’s posts
Want what’s new (early signals) → Latest posts
Want discussions on a specific topic → Search posts
Want the actual debate → Get item with ID (read comments)
Web Search Tools FAQ
There are several web search tools available on Hatz that grant AI real-time access to the Internet. The section below explains why and when each one should be used, and how to use them together for the best results.
What Is Each Web Search tool best at?
Firecrawl | When you need the actual page content (clean text/markdown/structured extraction) from known URLs or a site section |
Google Search | When you need broad coverage discovery (find lots of candidate pages quickly) |
Exa | When you want high-precision discovery (semantic relevance) and/or a citation-backed answer ( |
Tavily | When you want fast, LLM-optimized search results (quick discovery to feed a workflow or agent) |
Perplexity | When you want web-connected Q&A/research writing with sources (synthesis-first) |
How to choose when you can use 0 / 1 / 2 / 3 integrations
Number of Web Search Tools | When To Use Them | Recommended Tools |
0 | You do not need to access real-time info from the Internet |
|
1 | One step needed: find or read or synthesize |
|
2 |
|
|
3 | Full research pipeline |
|
Use 0 integrations (LLM Only)
Use none when:
The user already provided all necessary content, or
The task is purely transformation (rewrite, summarize provided text, classify, plan) and doesn’t require new external info.
Use 1 integration (pick one based on the bottleneck)
Firecrawl only: You already have URLs (or a site) and need to extract/normalize content to quote, summarize, or parse.
Google Search only: You only need a list of candidate sources (broad net, general discovery).
Exa only:
EXA_SEARCHwhen you want best-match sources fast (semantic discovery).EXA_ANSWERwhen you want a direct response with citations and don’t need to deeply ingest full pages.
Tavily only: You need quick discovery (fast SERP-like results optimized for LLM pipelines).
Perplexity only: You want a synthesized answer/research write-up grounded in the web, and don’t need to ingest raw pages yourself.
Use 2 integrations (most common: discover → extract, or answer → verify)
A) Discover → Extract (best for accurate downstream processing)
Google Search + Firecrawl: Broadly find pages, then scrape the best ones for real text.
Exa (Search) + Firecrawl: Find the most semantically relevant pages, then scrape for deep summarization/comparison.
Tavily + Firecrawl: Speed-first discovery, then scrape the selected URLs.
B) Answer → Verify / Quote
Perplexity + Firecrawl: Get a synthesized answer with sources, then scrape cited URLs to verify, pull exact quotes, or extract structured fields.
Exa (Answer) + Firecrawl: Get Exa’s citation-backed answer, then scrape the cited pages for auditability or deeper extraction.
Use 3 integrations (strong “research pipeline” patterns)
Pick one discovery tool + one synthesizer (optional) + Firecrawl:
A) Exa + Perplexity + Firecrawl (quality + synthesis + grounding)
Use Exa to find high-quality sources (or to answer with citations),
Use Perplexity to draft a coherent research narrative,
Use Firecrawl to extract/verify key pages and produce final, quote-accurate output.
B) Google Search + Perplexity + Firecrawl (breadth + synthesis + grounding)
Use Google to cast a wide net,
Perplexity to synthesize,
Firecrawl to pull authoritative pages and confirm details.
C) Tavily + Perplexity + Firecrawl (speed + synthesis + grounding)
Use Tavily for fast discovery,
Perplexity for the answer/write-up,
Firecrawl to extract supporting details and ensure you’re not relying only on snippets.
Simple rule of thumb
If you must read the page, use Firecrawl.
If you must find pages, use Google (broad) or Exa (high relevance) or Tavily (fast).
If you must produce a sourced narrative answer, use Perplexity or Exa_ANSWER.
Best reliability usually comes from: (Search) → Firecrawl extraction → final write-up.
Just because you turn on 1, 2, or 3 Web Search tools, the AI model you chose will not always automatically use all of them. That's why it's important to understand the tool calls and work them into your prompts / system prompts.
Use the Hatz Workshop Assistant to help with this!
Integrations - More Coming Soon
Salesforce Integration
Salesforce integration is the “look up and update CRM records from your workflows” integration: it lets you search, retrieve, create, and update key Salesforce objects (like Contacts, Leads, Accounts, Opportunities, Campaigns, Notes, and Tasks) so your CRM stays current without manual copy/paste.
Tool Calls Table
Find / query
Tool call | What it does (plain English) |
| Runs a Salesforce query (SOQL) like “find Contacts where email = X”. |
| Quick name-based search for Contacts (case-insensitive). |
| Runs a Salesforce report and returns its results (without editing the saved report). |
Retrieve (get a specific record)
Tool call | What it does |
| Fetch one Contact by ID. |
| Fetch one Lead by ID. |
| Fetch one Account by ID (optionally pick fields). |
| Fetch one Opportunity by ID. |
| Fetch one Campaign by ID. |
Create (make new records)
Tool call | What it does |
| Create a new Contact (requires LastName; AccountId optional). |
| Create a new Lead (usually requires LastName + Company). |
| Create a new Account (requires Name). |
| Create a new Opportunity (requires Name, StageName, CloseDate). |
| Create a new Campaign (requires Name). |
| Create a Note attached to an existing record (ParentId). |
| Create a Task (a to-do/follow-up) tied to a person (Lead/Contact) and/or a record (Account/Opportunity/etc.). |
Update (edit existing records)
Tool call | What it does |
| Update fields on a Contact. |
| Update fields on a Lead. |
| Update fields on an Account. |
| Update fields on an Opportunity. |
| Update fields on a Campaign. |
| Update a Note (title/body/privacy/parent). |
FAQ - How would the average person use it?
Most people use it to:
Look up a contact/lead and get the latest details
Log activity (create a Task, add a Note)
Create new records from forms or inbound requests (Lead/Contact/Account)
Keep the CRM clean (update phone/title/company, fix missing fields)
Pull a report result and summarize it for leadership
Is Salesforce integration read-only?
No—Salesforce integrations are typically read + write, depending on:
The specific tool call (query vs create vs update)
Your Salesforce user permissions and org rules (sharing, validation rules, required fields)
What’s a Salesforce “object”?
A category of records—like:
Contact (a person at an Account)
Lead (a potential customer not yet qualified/converted)
Account (a company/organization)
Opportunity (a deal)
Campaign (marketing initiative)
Task/Note (activity logging)
What’s a Salesforce ID?
A unique identifier for a record (often 15 or 18 characters). Many actions require it (e.g., “update this Contact by ID”).
What’s SOQL?
Salesforce’s query language (“Salesforce SQL”). You use it to search precisely (e.g., by email, domain, owner, date ranges).
Can it “do anything” in Salesforce?
Not automatically. It can only do what:
The integration supports (the available tool calls), and
Your Salesforce permissions allow (read/write on objects/fields), and
Your org rules allow (required fields, validation rules, automation, etc.).
Common questions people ask
“Find this person in Salesforce and tell me their Account, title, and last activity.”
“Create a follow-up task for next Tuesday and assign it to the account owner.”
“Update the contact’s phone and log a note about our call.”
“Create a new lead from this inbound form submission.”
“Pull a pipeline report and summarize trends.”
Can it handle duplicates or matching?
It can help, but you usually need to define the matching rule (most common: Email for people; Domain + Name for accounts). A good pattern is: search first → if not found, create → if found, update.
Prompts Using Salesforce (3 use cases, simple → advanced)
Use case 1 (simplest): Log a follow-up task after a meeting
Prompt:
“Create a Salesforce Task: subject ‘Follow up on pricing questions’, due date next Friday, related to this Opportunity (WhatId: [OPPORTUNITY_ID]) and assigned to [OWNER_ID]. Put my meeting notes in the description.”
Why it works: Tasks are the simplest, safest “CRM hygiene” action.
Use case 2 (mid-level): Find a contact and update details
Prompt:
“Search Salesforce for a Contact named ‘[NAME]’ and confirm the right record by email/domain. If found, update their title to ‘[TITLE]’ and phone to ‘[PHONE]’. Then add a note to the Contact summarizing: [NOTES].”
Why it works: “find → verify → update → log” mirrors how humans do CRM updates.
Use case 3 (more complex): Create or update lead from an inbound request (with guardrails)
Prompt:
“Use Salesforce to process this inbound lead: [paste details]. First, search by email to see if a Lead or Contact already exists.
If it exists: update missing fields and log a note ‘Inbound form submission’ with the form answers.
If it doesn’t exist: create a new Lead with Status = ‘Open’ (or the default), Lead Source = ‘Web’, and assign to [OWNER_ID].
Finally, create a follow-up Task due in 2 business days.”
Why it works: avoids duplicates and ensures a follow-up is scheduled.
Quick decision guide (one-liner)
Need to look something up → query (SOQL / name search)
Need to record activity → create a Task or Note
Need a new record → create Lead/Contact/Account/Opportunity (but expect required fields & validation)
Need to fix/complete data → update the record by ID
If you tell me your exact workflow (e.g., “inbound web form → CRM → follow-up task”), I can propose the safest “search-first” logic and the minimum fields you should require to avoid Salesforce validation errors.