Coming from SiteGPT?
SiteGPT provides a website support chatbot grounded in your content, with integrations and developer interfaces as well as answers. Whenever is a different fit for a custom workflow in your team tools: describe what to look up or change and supervise the results. AI-written replies can vary and consume runtime AI; building has weekly usage limits. The file is yours.
What you get on Whenever
Define the action your team needs
Choose integrations from Whenever’s 1,400+ catalogue or provide a documented API. Availability and actions depend on the service’s API, permissions, plan and reachability from the hosted runtime; a catalogue listing is not a guarantee that every operation or sign-in method is supported. Test the specific lookup or update before publishing.
Map, summary, run history and code
Review the workflow map, summary and run history. Run and step logs are retained for 7 days on Hobby and 30 days on Pro. The code can be downloaded and handed to a developer; that is not a guarantee that a reported action succeeded in an external service.
What one Whenever plan covers
One plan can run bots and other workflows within its live-workflow, run and AI-spend limits. At a limit, contact Whenever about a larger plan; external services may bill separately. Hobby permits one active workflow.
A concrete decision
Illustrative scenarios, not customer results or recorded runs.
Greg: answer an order question safely
Illustrative decision, not a recorded customer run: Greg wants a team-chat bot to look up an order ID and report shipping status, without issuing refunds. He specifies the allowed lookup and tests found, missing and ambiguous orders with authorised accounts. He checks the returned provider status and run history, not a model’s unsupported “done” message or source code. A failed lookup should be reported as unconfirmed, not as a shipped order. Choose a custom procedure for that cross-tool rule; keep a specialised helpdesk if its ready-made experience fits. Model-written replies can vary and use runtime AI; this example is not proof of a delivered customer action.
Jaunius: choose the trigger before pricing it
Illustrative calculation, not observed traffic: a 15-minute schedule requires Pro and runs 96 times a day, or 2,880 times in a 30-day month, even when nothing arrives. A supported inbound trigger counts the executions it starts: for example, WhatsApp Business messages or Discord slash commands, buttons and modals. Scheduled workflows can also call channel APIs. Jaunius chooses the trigger and tests duplicate events and tool permissions. Compare the subscription, run and AI allowances and external charges; 2,880 checks are not automatically 2,880 messages or a saving.
Message quotas and workflow runs differ.
SiteGPT says one message covers the question and AI reply, with model-dependent quota consumption. Whenever separately tracks executions and runtime AI spend. Compare the same conversation under the applicable plans.
SiteGPT
Plans with monthly message quotas
One message covers a user question and its AI reply
Plans set the number of chatbots, messages, training pages and seats. Self-serve plans start with a seven-day trial that needs a card.
SiteGPT says one message covers the user question and AI reply. A GPT-4.1 message uses the quota of ten GPT-4.1 Mini messages; this is a quota ratio, not a dollar comparison. Whenever separately tracks workflow executions and runtime AI spend. Compare the same conversation and actions under the applicable plans.
A run is one execution of one workflow, start to finish. AI building and editing are free within weekly usage limits; runs are metered separately.
A closer look
What changes when you describe the job in plain language and code runs it: to build, to run, and to keep.
Repeatable execution logic
SiteGPT
Model-generated replies use the sources and instructions you configure. SiteGPT also documents human replies; not every conversation message is a model-generated answer.
Whenever
The published code sets the bot’s allowed tools and execution limits. Review its rules and sample results; source-code inspection is optional. A model can choose wording and offered tools, so identical messages need not produce identical replies. Restrict permissions as well as the offered tools.
Runs per dollar
SiteGPT
SiteGPT plans set monthly message quotas and chatbot, page and seat limits. One message covers a user question and AI reply; GPT-4.1 uses ten times the quota of GPT-4.1 Mini. Confirm model, plan and add-ons before budgeting.
Whenever
Pro includes 10,000 runs per month; see current US dollar subscription pricing. Each workflow execution counts once within its call budget. Scheduled API checks count even when idle; supported inbound events count when they start a workflow. Runtime model usage draws on the plan’s included AI spend in runs. If you reach a run or AI-spend limit, contact Whenever about a larger plan; no self-serve overage price is published. AI building has separate weekly limits.
What you can express
SiteGPT
SiteGPT documents training sources, support integrations and developer interfaces including API, CLI and MCP. Do not assume a content-grounded chatbot cannot take actions.
Whenever
Describe the bot’s allowed actions and handoff rules in plain language. Its tools can use 1,400+ integrations or a documented API, subject to permissions, provider-plan support and network reachability from Whenever’s hosted runtime. Review and test the results before publishing.
Export and portability
SiteGPT
SiteGPT documents exports of conversation and lead data. Data export and moving the chatbot runtime are different requirements; confirm the latter before planning a migration.
Whenever
The bot is a TypeScript file: its loop, its tools and its rules. Read it, export it, hand it to a developer. Moving it to another runtime can mean adapting managed connections and triggers.
Debugging a failed run
SiteGPT
SiteGPT documents conversation history, filters and answer sources. Scale and higher plans include AI conversation insights that classify topics; confirm the plan needed for those analytics.
Whenever
Use the workflow map, summary and run history to inspect what happened, then ask for changes in chat and test representative messages. Run and step logs are retained for 7 days on Hobby and 30 days on Pro. Code is available when you or a developer want to inspect it.
Where the AI cost lands
SiteGPT
Model choice affects message quota: SiteGPT documents a 10:1 ratio for GPT-4.1 versus GPT-4.1 Mini. That is not a dollar comparison with Whenever.
Whenever
Optional model-written replies use the plan’s included AI spend in runs. At a run or AI-spend limit, contact Whenever about a larger plan; no self-serve overage price is published. Non-AI API operations do not require model inference, but external services may bill separately.
Where the bot talks
SiteGPT
SiteGPT provides a website widget. Slack, Messenger, Google Chat, Crisp, Freshdesk and Zoho SalesIQ are listed for Growth and above; Zendesk Messaging for Scale and above. WhatsApp is an add-on available across plans.
Whenever
Supported inbound events include Slack app events, Telegram bot updates, WhatsApp Business messages through a Meta app, and Discord slash commands, buttons and modal submissions. A scheduled workflow can also call the channel API; each scheduled run counts even when it finds nothing. This does not promise push delivery of ordinary Discord channel messages. Whenever does not host a website chat widget.
Building from your coding agent
SiteGPT
SiteGPT documents dashboard setup and API, CLI and MCP interfaces. Compare the interface and permissions your intended integration needs.
Whenever
Connect Claude Code, Cursor or Codex to the Whenever MCP server. Your coding agent can write and test the workflow and save it as a draft; complete Setup and publish after reviewing the result.
When to keep SiteGPT
SiteGPT
Consider SiteGPT for a website widget grounded in your pages and documents with human handoff. Its plans list manual refresh on Starter and Auto Refresh of source data on Growth and higher, with cadence varying by plan; files from connected apps are resynced by hand.
Whenever
Consider Whenever for a custom cross-tool procedure in a team-chat workflow. Compare the action and handoff with SiteGPT’s own integrations; Whenever does not provide a built-in knowledge-retrieval product.
Sources and scope
Official sources captured on the date shown below. These describe published features and billing rules, not independent performance tests. Check the linked documentation for current modes, plans and limits. No competitor dollar price is asserted.
Checked
Support jobs to start from
Starter briefs for related jobs, not recorded runs. Connect authorised accounts, confirm trigger and plan requirements, then review and test the workflow before publishing.
Build this workflowSpot expansion signals in support ticketsSends the account owner a written read on every support ticket that mentions buying or growth.Build this workflowFlag renewal risk on upcoming Salesforce accountsReviews Salesforce accounts renewing in the next 90 days with their open tickets and posts a written risk note to Slack.
Build this workflowSend a pre-call brief before every customer callAssembles account details, deal history, and open tickets into one brief and sends it to the call owner before the meeting.Build this workflowTurn support tickets into upsell signalsReads each new Zendesk ticket for upsell and adoption gaps, writes the finding on the HubSpot account, and tells the CSM.Build this workflowAdd customer context to every new Zendesk ticketLooks up each new ticket's requester in HubSpot and adds their company, plan, open deals and owner as an internal note.Build this workflowStart with one support workflow.
Describe what it should look up and change. Whenever writes it as code you can read and keep.
Pick the question that needs a lookup
Choose a lookup or update and record its permissions, handoff rules and expected destination. Compare it with the SiteGPT integrations you already use rather than assuming an action is impossible there.
Describe what should happen
Say which app has the answer and what the bot may change. Whenever writes it; review the rules and connect each app.
Test it with real messages
Send it the questions your customers ask, the odd ones included, before you point anyone at it.