PRICING
Free on your machine. Metered in the cloud.
Run the desktop app for nothing and keep every record on your own computer. Pay a flat monthly fee only when you want your pipelines and code saved in the cloud, and buy compute credits for the cloud work you actually do. Credits roll over — you never lose what you paid for.
Local
$0forever
Everything runs on your own machine, and every record stays there.
- Desktop app for macOS, Windows, and Linux
- Unlimited pipelines, blocks, and runs
- Pipeline data stored in PostgreSQL on your computer
- Collection and pipeline jobs run on your hardware — no metered compute
- Your own model keys in the encrypted Key Vault
- Every generated file downloadable
- Nothing saved to the cloud — this machine is the only copy
- Maintained sources not included
No account and no card required.
Basic
$10/ month
Your pipelines and their generated code, saved in the cloud and reachable from any browser.
Start Basic- Everything in Local
- Pipelines, blocks, and generated code saved to your workspace
- Build and edit from any browser — no install
- Version history for every pipeline and orchestration
- Server-side schedules and orchestrations that run with the browser closed
- Managed storage, plus MCP servers over your workspace tables
- Cloud runs draw from prepaid compute credits that never expire
- Maintained sources not included
The subscription covers saving and scheduling your work. Cloud runs draw separately from your credit balance.
Pro Maintained sources
$29/ month
Every source in the Source Library, kept current and fixed for you.
Start Pro- Everything in Basic
- The full Source Library — every maintained source, connectable from any pipeline
- Sources monitored on a schedule and repaired as new versions
- Validation evidence and version history for each source
- Request a source you need added to the library
- Higher included managed storage
- Cloud runs draw from prepaid compute credits that never expire
Sources you build yourself work on every plan. Pro is for the ones we keep working.
↓Collection compute
Fetching pages from the web
Credits are drawn per page that comes back with content. Collection is the more expensive of the two meters: reaching a live site means proxying, retries, and — on protected sites — a full browser, all of which we carry for you.
- Standard page$0.20 / 1,000 A plain fetch. Most listings, feeds, APIs, and article pages. 1 credit collects 5,000 of them.
- Rendered page$1.80 / 1,000 Sites that need a real browser or anti-bot handling. The rate for a source is shown when you build it, before it ever runs.
For exampleA source collecting 200 pages a day at the standard rate draws about 1.20 credits a month. Failed and rate-limited fetches draw no credits.
⚙Pipeline compute
Running the blocks in your pipeline
Credits are drawn for the time your pipeline actually executes — transforms, joins, enrichment, deduplication, and writes to storage. This is ordinary CPU and memory, so it costs a fraction of collection.
- Compute time$0.15 / hour Metered by the second on 1 vCPU and 2 GB of memory. Idle time between runs draws nothing.
- Managed storage$0.20 / GB-month Beyond the allowance included with your plan. Only what your tables and object stores actually hold.
For exampleA pipeline that takes 90 seconds draws about 0.004 credits a run — roughly 0.11 credits a month running every morning.
Credits are spent only when a run executes in the cloud. Nothing expires, and nothing resets at the end of the month.
PREPAID, NEVER EXPIRING
Both meters draw from credits you already bought.
Cloud compute is prepaid. You buy credits when it suits you, and collection and pipeline runs draw them down at the rates above. One credit covers one dollar of metered compute, so every rate on this page reads the same in credits.
There is no monthly compute bill to be surprised by, and no allowance to spend before it disappears.
- Buy what you needTop up in any amount from your workspace. Buy once for a one-off backfill, or in larger blocks if you run daily.
- Credits roll overUnused credits never expire and are never reset at the end of a month or a billing period. What you bought stays yours until a run spends it.
- Auto top-up when lowSet a threshold and an amount. When the balance falls below it, the next top-up is added automatically, so an overnight schedule never stalls halfway through.
- Local runs spend nothingA pipeline executed on your own machine draws no credits at all — the balance sits untouched.
ZERO CREDITS SPENT
Or run both meters at zero.
Neither meter is compulsory. The desktop app executes the same pipelines on your own machine: collection goes out over your network, blocks run on your CPU, and results land in a PostgreSQL database on your computer. No credits are drawn.
This works on every plan. Keep a Basic or Pro subscription for saving, scheduling, and maintained sources, and still run the heavy work at home.
- Install the desktop appmacOS, Windows, or Linux. It checks for Python 3 and PostgreSQL on first launch.
- Pick the local runtimePoint a pipeline at your machine instead of managed compute. The graph and blocks are unchanged.
- Add your model keyAI blocks unlock as soon as a provider key is in the encrypted Key Vault.
- Run itPages are fetched from your network and rows are written to your own database. Your credit balance does not move.
Run summary
Credits spent: none. Model calls are the only cost in a local run, and they are billed to you by your provider — never through Pipechat.
BRING YOUR OWN KEY
Every model call is billed by your provider, not by us.
Pipechat does not resell model tokens and does not add a markup. Every call to a language model — asking the chat to build a pipeline, an LLM Transform block classifying a record, an Embeddings block vectorising text — uses a key you connect in the encrypted Key Vault, and your provider bills you directly at their published price.
That is true on all three plans, in the browser and on the desktop. It also means you choose the model: swap providers, use a cheaper model for bulk enrichment, or point a block at something you host yourself. Nothing on this page changes when you do.
DETAILS
Frequently Asked Questions
- What counts as a page?
- One successful fetch of one URL — a listing page or an individual item. Redirects, retries, rate-limited responses, and failed fetches draw no credits.
- How do I know which rate a source is on?
- Pipechat samples real pages when it builds a source, so it knows before you ever run it whether the site needs a browser. The rate is shown on the source, and again in the run estimate.
- Do credits expire?
- Never. Credits roll over month after month and are not reset at the end of a billing period. A quiet month costs only the subscription, and the balance is still there when you come back.
- What happens if I run out mid-run?
- With auto top-up on, the next top-up is added and the run continues. With it off, the run pauses at the block it reached and resumes once you add credits — collected data is not discarded.
- Is there a minimum purchase?
- No, and no commitment. Buy credits when you need them; since nothing expires, buying more only means buying less often.
- Can I use maintained sources on Basic?
- No — the Source Library is the Pro plan. Sources you build yourself work identically on every plan, including Local.
- What happens if I cancel?
- Your pipelines stop running in the cloud, and every generated file stays downloadable. Move the project to the desktop app and it keeps running on your own machine for nothing.
- Do I need a subscription to run locally?
- No. The desktop app is free and unlimited on its own. A subscription adds cloud saving, scheduling, and — on Pro — maintained sources.
Start where it costs nothing.
Run the desktop app today and move to the cloud when saving and scheduling are worth $10 a month.
Create your account