First: identify where you’re logging in
IF you’re trying to manage tenants/users/org settings (MSP/admin tasks)
THEN you should be using admin.hatz.ai (admin portal).
IF you’re an end user chatting/using tools in your org (NOT an MSP)
THEN you should be using your org’s subdomain (your unique Hatz URL), not admin.hatz.ai.
Source: “Access Restricted” note in the error guide + login troubleshooting context.
“Hatz doesn’t load” / pages error / chats & workflows won’t load
IF the app won’t load (chat history/workflows not loading) and your internet is fine
THEN whitelist Hatz domains on your network/security stack:
Preferred:
*.hatz.ai(if wildcards supported)Or explicitly allow:
db.hatz.aiapi.hatz.aiai.hatz.aiadmin.hatz.ai[your-subdomain].hatz.aiauth.hatz.aiassets.hatz.ai
Source:
“Access Restricted”
IF you see “Access Restricted”
THEN confirm you’re logging into the correct place:
admin.hatz.ai is only for your org’s Hatz AI administrators (MSP/admin users)
End users should use their unique org subdomain URL
Source:
Login, invite, and password reset problems (admin vs customer users)
Start by identifying the user type:
Admin user (logs in via admin.hatz.ai) — password reset
IF an ACTIVE Admin user needs a password reset
THEN in admin.hatz.ai go to Settings → Users and Roles → (…) → Reset Password.
Admin user — invite wasn’t received/expired
IF an INACTIVE Admin user didn’t get the invite / invite expired
THEN in admin.hatz.ai go to Settings → Users and Roles → (…) → Resend Invite.
Customer/Tenant end user (logs in via your subdomain) — password reset / invite issues
IF an ACTIVE Customer user needs help logging in
THEN in admin.hatz.ai go to Tenants → select tenant → (…) → Resend Invite (sends reset/setup instructions).
IF a PENDING Customer user didn’t receive the invite / it expired
THEN in admin.hatz.ai go to Tenants → select tenant → (…) and choose either:
Resend Invite, OR
Generate Access Link (useful when email is getting caught in spam)
Source (all of the above):
Chat is slow / “taking too long to respond”
IF chat is taking too long to respond
THEN
Refresh the page
If still slow, switch to a different LLM/model
Source:
(Helpful context on model selection UI/attributes:)
“This chat used to work, now outputs are worse”
IF an old/long-running chat stops giving good outputs
THEN assume you’re hitting the model’s context window limits; move to a fresh chat (or reduce what you’re asking it to consider).
Source (context window note):
Outputs differ from ChatGPT / OpenAI website results
IF you’re using an OpenAI model in Hatz but results differ from ChatGPT on openai.com
THEN check these common causes:
You may not be using the exact same model version
Different system prompts / default instructions
Natural randomness
Different generation settings (temperature / max tokens / etc.)
Different context handling
Rate limits/performance differences
UI “enhancements” on the website that aren’t part of raw API usage
Source:
File won’t upload
IF a file won’t upload even though it’s under 30MB
THEN it may still exceed the token limit for the specific LLM you selected; try a different model (or reduce the document size/scope).
IF a file won’t upload even though it’s under 30MB
THEN it might not be an accepted file type. We can evaluate these issues on a case by case basis.
Source:
Tools / web search / integrations aren’t behaving as expected
IF web search or an integration isn’t doing what you expect
THEN
Confirm you selected an LLM that supports tool use (some models don’t)
Use Inline Tool Calling to see what the AI actually searched/used and what came back (helps debug prompts + tool behavior)
If you need different results, pick a different search tool:
Google Search = fast fact-finding
Firecrawl = deep scrape of a specific site
Perplexity = best for current events/news
Sources:
Tool selector guide: https://docs.hatz.ai/en/articles/11900262-ai-tool-selector
Inline tool calling: https://docs.hatz.ai/en/articles/11639134-inline-tool-calling
Model selector (attributes like “Tool Use”): https://docs.hatz.ai/en/articles/11501162-ai-model-selector
“We can’t see a model” / “too many models” / org-wide model defaults
IF users can’t find a model they expected (or you want to standardize model choice)
THEN check org-level settings:
Set a Default LLM (shows first; becomes default in chat/workshop)
Disable LLMs (hide specific models or whole vendors)
Source:
Credits / usage limit questions
IF the issue is “credits, quotas, or usage limits”
THEN use the official credit usage FAQ (most accurate source, updated policy):
(Background explainer in docs collection:)
“I’m not sure if I should build an App, Workflow, or Agent”
IF you need multiple steps / multi-output / run history
THEN build a Workflow.
IF you need a reusable chatbot persona that can use tools/integrations and sources every time
THEN build an Agent (also usable directly in Chat).
IF you need a single AI “app” (prompt + inputs), including API usage
THEN build an App.
Source: https://docs.hatz.ai/en/articles/12281902-workshop-faq
Workflow troubleshooting
“My workflow step can’t see the previous step’s output”
IF you can’t insert a prior step output into a later step
THEN confirm the steps are connected (connections are required to insert earlier outputs into later steps).
Source: https://docs.hatz.ai/en/articles/11328145-workflows-beta
“Step 2 is ignoring Step 1 / doesn’t use the earlier info”
IF Step 2 runs but behaves like it didn’t get Step 1 context
THEN explicitly insert Step 1’s Output into Step 2’s prompt (treat each step like a distinct automation and include the needed context).
Source: https://docs.hatz.ai/en/articles/11328145-workflows-beta
“My workflow web scrape didn’t crawl the whole site”
IF you expected the workflow’s Web Scraper Step to click links / crawl the domain
THEN note it only scrapes the single page URL you provide (it does not crawl the whole site).
THEN either (a) provide the exact URL you need, or (b) redesign steps so earlier outputs produce the right URL for the scraper step to use.
Source: https://docs.hatz.ai/en/articles/11328145-workflows-beta
“I need to know which step used which model / credits / why it’s expensive”
IF you need to debug cost, latency, or model choice inside a workflow
THEN use Workflow Run History to inspect:
credits used (total + per-step)
time spent (total + per-step)
model used per step
step inputs + outputs
Source: https://docs.hatz.ai/en/articles/11328145-workflows-beta
“I want a clean deliverable (Word/PDF) from a workflow step”
IF you want to share outputs as documents
THEN export any workflow step to PDF or Word.
IF you want branded/structured Word docs
THEN use a Document Processing Step with a .docx template and map placeholders (e.g., {{client_name}}) to workflow data.
IF you need to fill an existing PDF form
THEN upload a fillable PDF and map workflow values to its form fields.
Source: https://docs.hatz.ai/en/articles/12332454-ai-document-processing-steps-and-exports-in-workflows
“Where do I find Workflows / Automations again?”
IF you can’t find where to build/manage workflows
THEN go to the Workshop and look under Automations (Workflows are inside Workshop).
Source: https://docs.hatz.ai/en/articles/11328145-workflows-beta
(Workshop navigation/context): https://docs.hatz.ai/en/articles/12572765-the-hatz-workshop
Agent troubleshooting
“My agent can’t use an integration / keeps saying it can’t access X”
IF an agent can’t use tools/integrations
THEN verify both:
the tool/integration is actually enabled for the agent (Tools & Integrations selection), and
the model supports tool calling (agents that need tools must run on a tool-calling capable model).
Source: https://docs.hatz.ai/en/articles/12602274-agents
“I can’t add more tools to my agent”
IF you hit a limit adding tools
THEN note agents can select up to three tools.
THEN either swap tools, or create multiple specialized agents and switch agents when needed.
Source: https://docs.hatz.ai/en/articles/12602274-agents
“The agent isn’t following the persona/rules I intended”
IF the agent’s behavior is inconsistent
THEN tighten the System Prompt (Instructions): define role, context, tasks, constraints, and output format (so it runs every time).
Source: https://docs.hatz.ai/en/articles/12602274-agents
“I built an agent but my team can’t see it”
IF others can’t access your agent
THEN check sharing settings—agents default to Private; enable Share with Organization to make it available org-wide.
Source: https://docs.hatz.ai/en/articles/12602274-agents
“How do I use the agent in Chat?”
IF you created an agent but can’t figure out how to talk to it in Chat
THEN use the agent selector in Chat (robot icon) and select your agent from the list.
Source: https://docs.hatz.ai/en/articles/12602274-agents
Integrations + tool calling troubleshooting (Chat + Workflows + Agents)
“I don’t see integrations in the Tool Selector / can’t connect them”
IF integrations aren’t appearing or you can’t connect them
THEN check where you’re trying to connect from:
Integrations can be accessed in Chat/Workflows via the Tool Selector
They can also be connected from the Integrations tab (noted as being in Workspace on the Customer Dashboard in the doc) Source: https://docs.hatz.ai/en/articles/11900262-ai-tool-selector
“Integrations are in beta / missing for a tenant”
IF you need beta integration access and don’t see the integration
THEN enable beta features for the right place:
IF you are an MSP Admin (Admin Dashboard)
THEN enable Beta Features for the tenant in the Admin Dashboard (toggle Beta Features, refresh).
Source: https://docs.hatz.ai/en/articles/11900366-how-to-enable-access-to-beta-features-in-the-admin-dashboardIF you are a Client Admin (Customer Dashboard)
THEN enable Enable Beta Integrations in Org Settings (then try again).
Source: https://docs.hatz.ai/en/articles/11905828-how-to-enable-access-to-beta-features-in-the-customer-dashboard
“The AI used the wrong tool (or no tool)”
IF results are stale or it didn’t search/scrape when it should have
THEN either:
manually select tools via the Tool Selector (up to three), or
rely on Automatic Tool Selection, then verify what it picked. Sources:
Tool Selector: https://docs.hatz.ai/en/articles/11900262-ai-tool-selector
Automatic Tool Selection: https://docs.hatz.ai/en/articles/12939487-automatic-tool-selection
“I want to see what the AI actually searched/called”
IF you’re debugging a workflow/agent/integration and want proof of what happened
THEN use Inline Tool Calling to view:
what it searched
which tools it used
the tool responses
Source: https://docs.hatz.ai/en/articles/11639134-inline-tool-calling
Quick “build & deploy” gotchas (Workshop items)
IF you duplicated a community template but can’t find it / edit it
THEN check My Items in Workshop (duplicating creates your editable copy; items start private).
Source: https://docs.hatz.ai/en/articles/12572765-the-hatz-workshop
IF you published something but the wrong people have access
THEN re-check publishing/sharing scope (private vs org vs client tenants).
Source (apps publishing, relevant pattern for access scoping): https://docs.hatz.ai/en/articles/10716519-publishing-apps
When to escalate
IF you’ve followed the branch above and it still fails
THEN contact Hatz support: [email protected] (listed in Hatz docs).
Source:
