# Whenever > Whenever is the platform for AI workflows: deterministic flows, AI steps and bounded agent loops. A coding agent or Whenever’s own builder writes a TypeScript workflow, and Whenever hosts, triggers and runs that reviewed code, with managed secrets, integrations and run history. Orchestration is code; a model is called only by a step the workflow declares. Metered per run; free AI building and editing have weekly usage limits. ## The platform for AI workflow automation. Build, deploy, and run deterministic flows, AI steps, and full agents in one place. [Deploy a workflow from Claude Code, Cursor or Codex](https://www.whenever.dev/deploy) Used by teams behind Wix, Symphony, Wixel, Base44, Tabless and Brik. ## For coding agents - To deploy: read https://www.whenever.dev/skill.md, then connect the MCP server https://mcp.whenever.dev (Claude Code: `claude mcp add --transport http whenever https://mcp.whenever.dev`). - Install the guide as a Claude Code skill: `mkdir -p ~/.claude/skills/whenever-deploy && curl -fsSL https://www.whenever.dev/skill.md -o ~/.claude/skills/whenever-deploy/SKILL.md`. - There is no CLI and no import format: rewrite the existing task as a workflow module, save a draft over MCP, and publish. - The owner enters every secret in the browser Setup page; no MCP tool accepts a credential. - A policy pass is not a build, a build is not a publish, a publish is not an activation, and none of them is a successful run. ## Build it anywhere. It runs whenever. ### Build on Whenever: From prompt to production workflow - **Describe it.** Describe it in plain words. It asks only what it needs to know. - **See the flow.** Every step, easy to follow. Connect apps in one click. - **Fix it in chat.** Something broke? AI reads the failure and fixes it with you. - **Watch it succeed.** Run completed with no new items recorded: no rows added or Slack messages sent. - **See the code.** Nothing hidden, nothing locked in: readable code a developer can take over. - **Build another.** Start the next one from a template, then change anything by asking. ### Build with your coding agent: Your coding agent writes it. Whenever runs it. Ask Claude Code, Codex or Cursor to turn the task your agent repeats into a workflow. It writes the code, saves a draft and deploys it to Whenever through our MCP server. You connect the accounts and decide when it goes live. MCP server: `https://mcp.whenever.dev` — [MCP Setup Guide](https://www.whenever.dev/mcp) Ask your coding agent: > Every weekday at 8am, find yesterday’s Stripe refunds over $200 and post a summary to #finance with each customer and the reason. > Use https://www.whenever.dev/skill.md ### Deploy an existing workflow: Bring the workflow you already have. Already have it somewhere else? An n8n workflow, a Claude task or a ChatGPT scheduled task: give it to your coding agent with our skill, and it reshapes it into a Whenever workflow and deploys it. ``` Turn this into a Whenever workflow and deploy it. Follow www.whenever.dev/skill.md ``` [How the skill works](https://www.whenever.dev/deploy) ## Stop paying your agent to rediscover the same workflow. AI is brilliant at building. Wasteful at repeating. Turn your agents into code: same result, a fraction of the cost, and the same steps every time. - **Same result, every run\*.** For invoices, refunds and payouts, a 1% error rate is too much. Whenever makes the same decision every time. - **Faster\*.** Whenever runs your workflow in milliseconds. Nothing waits for a model to think it through again. - **Cheaper\*.** Whenever uses AI once, to write your workflow. After that, every run is almost free. - **AI where it helps.** Need a reply written, a PDF read or a judgment call? Add an AI step in Whenever, only where you need it. \*For the steps that run as code. Add an AI step and that step can vary, and uses your plan’s AI spend in runs. ## Don’t start from a blank prompt. Pick a starting point, connect your accounts and review what Whenever builds before publishing. [Browse all templates](https://www.whenever.dev/templates) ## Any API, even the one nobody supports.\* 1,402 integrations in the catalogue. Not in the list? Name the service and Whenever reads its API documentation, tells you which pages it used, and writes the integration. Nothing published? Paste the docs you do have. \* The API must be reachable from Whenever’s hosted runtime. You still need the provider’s required credentials, permissions and plan; private-network-only services are not automatically accessible. [Browse all integrations](https://www.whenever.dev/integrations) ## Whenever it triggers, it works. Triggers, credentials, run history and versions are built in. There's no server for you to run. - **Triggers.** Webhooks, schedules, app events and on-demand runs. - **Managed credentials.** You enter them once in the browser. Nothing to host, and your coding agent never sees a secret. - **Step-level run history.** Every step’s input, output and error, kept 7 days on Hobby and 30 on Pro. - **Safe deploys.** Every save is a version. Production keeps running the one you turned on until you switch, and any version exports as TypeScript you own. - **Told when it breaks.** While you build, the builder reads the logs and helps you fix it. In production, you get an email when a run fails, and fix it in the app or from your coding agent. ## Free to start. Pay per run. Start with one free workflow, no credit card. Upgrade when you need more workflows or runs. No per-step or per-item counting: one run is one run, however many steps it has. - **Hobby: $0/mo.** 100 runs per month, Includes $1 of AI usage for AI steps and agents. One active workflow, no credit card. - **Pro:** 10,000 runs per month, Includes $10 of AI usage for AI steps and agents. [Find the right plan for you](https://www.whenever.dev/pricing) 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. ## Look inside the black box. Node builders hide complex logic in Code nodes that nobody can read. Whenever writes the whole workflow as code, so every step reads in plain words. ### Or bring one over and stop tracing it node by node Describe the automation you already have and Whenever rebuilds it as one file you read top to bottom instead of a canvas you follow with your finger, testable before it goes live. - [Coming from n8n](https://www.whenever.dev/vs-n8n): Describe what your node graph does and Whenever rebuilds it as a module you read top to bottom. - [Coming from Zapier](https://www.whenever.dev/vs-zapier): Describe your Zap’s triggers and actions and get a workflow you can edit, test and keep. - [Coming from an agent tool](https://www.whenever.dev/deploy): Turn the job your agent repeats into a workflow: the steps become code, and a model is called only where a step asks for one. ## The things people ask first ### Who builds Whenever? Wix.com Ltd, the Nasdaq-listed company (ticker WIX). Whenever is one of its products, not a site built on the Wix website platform. ### How do I deploy from Claude Code or Codex? Follow the setup on [whenever.dev/mcp](https://www.whenever.dev/mcp): it shows how to connect Claude Code, Codex or Cursor and deploy your first workflow. ### Can Whenever run AI agents? Yes, as a step in a workflow: an agent loop with the tools you give it and a capped number of turns, within a run of up to an hour. It sits next to fixed steps and single AI calls in the same workflow. Whenever is not an always-on chat agent. ### Do I have to write code? No. You describe what you want in plain language and Whenever writes the workflow. The code is there when you want it, to read, to review, or to edit, and ignoring it is a perfectly normal way to use Whenever. ### What does "workflow-as-code" mean? AI writes your automation once, as a typed TypeScript module, and every run executes that code. A workflow can be fixed steps only, call a model in one step where judgment is needed, or run an agent loop inside a step. The fixed steps follow the same rules on every run; AI steps and agent loops can vary between runs and draw on your plan’s included AI spend in runs. ### If I add an AI step, does every run still give the same result? Not for that step. Everything around it runs the same way every time, but an AI step asks a model, and a model can word its answer differently, or decide differently, from one run to the next. The rest of the workflow can check that answer before anything acts on it. An AI step also draws on your plan’s AI allowance on the runs where it happens. ### What happens when an API changes? Nothing changes shape under you: the workflow keeps running the code it was given. When a call starts failing, your run history shows which step broke, and you can ask Whenever to rewrite that step against the new API. ### How do triggers work? A workflow runs on a webhook, on a cron schedule, or on demand. Credentials are managed for you and there is nothing to host, so there is no server to babysit between runs. ### What happens if a run fails? Are there retries and alerts? A failed run is not retried automatically. While you build or test, the builder shows the run’s logs and helps you fix the workflow. In production, you get an email when an automated run fails, at most one per workflow per 24 hours, and you fix it from the logs, in the app or with your coding agent. ### Can I see the history of past runs? Yes. Every workflow has a run history: when each run happened, what started it, which version ran, and whether it succeeded. Open a run to see each step, how long it took, where it failed and what it returned. Runs and their steps are kept 7 days on Hobby and 30 days on Pro. ### Are my workflows versioned? Can I roll back? Every saved change becomes a new numbered version, and your live workflow keeps running the version you turned on until you turn on the new one. Each run shows which version it used. There is no one-click rollback yet: to go back, describe the change you want undone and Whenever edits the workflow again. ### How long is run history kept? Run history is kept for 7 days on Hobby and 30 days on Pro. Those are the retention periods listed on the pricing page. ### Can I export my workflows? Yes. A workflow is a TypeScript module, not a diagram locked inside a database, so you can take it with you. ## Your first workflow is one sentence away. Deploy your first workflow from your coding agent, or describe it in plain words. One stays live for free, no credit card. [Build with your coding agent](https://www.whenever.dev/mcp)