One price a month for what the plan includes. Only the two things that vary with use - what the AI costs and how long your scripts run - are metered, and every plan says up front what happens when you reach the end of them.
The AI allowance is spent at cost: the model provider's standard token price, with no markup. It also goes further here than it would elsewhere, because most steps never reach a model: a script runs wherever a script is enough, and you choose the model on the steps that do need one. Your admin can set a spend cap that stops the work rather than surprising you. How that works, in five minutes →
A 30-day demo workspace, given away rather than sold. One AI build and one AI analysis per person per day, no custom scripts, no triggers, and a shared cluster staff controls.
Premium Solo
LLM spend and compute included, and the excess billed rather than refused. Webhook and email triggers.
All features but with your own LLM tokens
Everything included.
Everything above is a number somebody chose. If none of them are your numbers, they are all negotiable - tell us what the work looks like and we will price that instead.
Every line of every category above, with what each one actually means - then the figures, and the four things that differ between plans.
| Per month | Demo | Premium Solo | Business | Bring your own LLM | Enterprise | Custom |
|---|---|---|---|---|---|---|
| What it costs | ||||||
| Base cost | 0 | 80 | 400 | 650 | 1200 | by arrangement |
| AI spend over the allowance As a share of the standard token price: 100% is at cost. | work stops | 200% | 200% | billed by your provider | 130% | by arrangement |
| Each extra minute of compute | work stops | 0.0500 | 0.0500 | work stops | 0.0300 | by arrangement |
| Setup run with you by PipeDuck Every plan gets the same guided walkthrough in the app; this is who works through it with you. | by arrangement | |||||
| AI | ||||||
| AI assistant for workflow build Describe what you want and get the graph back - the steps wired together, the arguments filled in, ready to edit. | ||||||
| AI assistant for script writing Draft a script and its unit tests from a prompt, against the shared library and the functions your workspace already has. | ||||||
| AI review before a script is approved Every new version is read, summarised and classified safe or unsafe on its way to the approver, so nobody signs off code they have not understood. | ||||||
| AI analysis when a run goes wrong Ask a failed run what happened: the cause, the step it broke on, and what to change - without reading a log. | ||||||
| AI steps inside a workflow A model as a step of the graph, given exactly the tools you hand it, with the cost of the call recorded like any other. | ||||||
| LLM chat with the tools you choose Talk to a model and hand it your own functions, per turn, with what each message cost shown as you go. | ||||||
| Automation | ||||||
| Every node type on the canvas Script, AI, condition, human approval, split and merge steps wired together - no code required, and never a dead end for developers. | ||||||
| Shared scripts A library of ready-made functions your own scripts import and build on, growing with every release. | ||||||
| Manual and scheduled runs Start a workflow yourself, or put it on a schedule - on every plan, before any trigger is bought. | ||||||
| Retries that do not need you A step that hit a timeout or a dropped connection is tried again on a backoff, against the budget the workflow already has. | ||||||
| Human approval steps, with a queue Stop a workflow on somebody, tell them, and let them approve or reject it from one page listing everything waiting on them. | ||||||
| Your own databases, queried by name Connections your admin registers once; a script asks for them by name and never carries a connection string. | ||||||
| Visibility | ||||||
| Alerts when a run fails An email or a webhook of your own the moment a run fails, stalls or is refused - rate-limited per rule, so a system that is down does not send sixty of them. | ||||||
| Dashboards Drag-and-drop grids watching your workflows live, built without code and shared with whoever needs them. | ||||||
| What the month is costing, as it happens AI spend and compute measured against the allowance while the month runs, not on the invoice afterwards. | ||||||
| Control and flexibility | ||||||
| Every run kept Inputs, outputs, logs, duration and cost recorded for each execution, and searchable afterwards. | ||||||
| The AI models you choose Your admin approves which models the workspace may use; every AI step and every chat picks from that set, and nothing else is reachable. | ||||||
| A spending ceiling of your own Set a monthly cap under the plan's own, and divide the pool between your workspaces - past either one the platform stops rather than spends. | ||||||
| Security | ||||||
| Platform security, built in Scripts run in their workspace’s own cluster and virtual environment, reach the network only through a filtered egress with no route to the internal network, and never run at all before a version has been approved. | ||||||
| Security that flexes to what you need Turn it up or down per workspace: two-factor sign-in, single sign-on against your own identity provider, approval of every individual function a script may call, and the exact list of addresses it may reach. | ||||||
| Approvals and versioning Nothing runs unapproved, every workflow, dashboard and script edit is attributed, and any of them is instantly reversible. | ||||||
| Access scoped to a team Workspaces nested to match your organisation, with a workflow or a script grantable to one team rather than to everybody in it. | ||||||
| Included allowance | ||||||
| LLM spend What the platform may spend with the model providers on your behalf, per month, at their standard token prices with no markup. | 25 | 15 | 100 | billed by your provider | 400 | by arrangement |
| Compute Total execution time across every script and workflow, per month. | 600 min | 100 min | 750 min | 2000 min | 4000 min | by arrangement |
| Storage How much your uploaded, received and produced files may hold at once. Your own retention setting is what keeps the total under it: files older than it are deleted automatically, and their names stay readable afterwards. | 2000 MB | 100 MB | Unlimited | Unlimited | Unlimited | by arrangement |
| Seats People who can sign in. | Unlimited | 1 | 25 | 150 | 250 | by arrangement |
| Support requests Questions that reach a person at PipeDuck, per month. The Support AI is on every plan and is not counted here - it answers immediately, and it can see your own setup, workflows and failed runs as well as the documentation. | Support AI only | Support AI only | Support AI only | Support AI only | Support AI only | by arrangement |
| Clusters Separate execution pools, each with its own package set. | 1 | 1 | 1 | 4 | 5 | by arrangement |
| Workers How many runs a cluster carries at once. | 2 | 2 | 6 | 16 | 20 | by arrangement |
| Workspace depth How deep you may nest workspaces inside the company, the company itself being level 1. | 2 | 2 | 3 | 4 | 4 | by arrangement |
| Added by this plan | ||||||
| Webhook triggers | by arrangement | |||||
| Email triggers | by arrangement | |||||
| Cluster routing | by arrangement | |||||
| Premium AI analysis | by arrangement | |||||
| Premium AI support | by arrangement | |||||
| Premium support | by arrangement | |||||
| AI discussion | by arrangement | |||||
| Support AI | by arrangement | |||||
Accounts are opened by the people running this instance, not by signing up - so a plan starts with a conversation rather than a card. Once you are on one, your workspace admin can split the allowance between teams and set a spending cap of their own underneath it.
Write to us and we will open the workspace: contact@pipeduck.com. Tell us roughly how many people and what you want to automate, and the answer comes back with a plan in it.