Open-source Clay alternatives
There is no open-source Clay, and searching for one is the wrong frame. Clay is three things in one tab: a table, a waterfall that calls data providers, and an AI column that researches each row. Open source replaces the first two very well. It does not replace the data.
Reviewed by Théo Pascard, Co-founder · Updated November 2026
Quick answer
Rebuild Clay as a stack, not as a single tool. n8n runs the waterfall and the dedupe that Clay's columns do, at the same cost for a 3-step or a 30-step workflow. Supabase is the table: one Postgres base of every company and lead already sourced, so nothing gets paid for twice. Playwright reads the sites that a plain request cannot, for research and personalisation. We run all three in production. What you still pay for is the data itself, the email finders and databases, which is most of what a Clay credit buys anyway.
The shortlist
| Tool | Best when | Starting price | Pricing | Market rating | Our experience |
|---|---|---|---|---|---|
| n8n | You want Clay's waterfall and routing without paying per row | Free, then €20/mo | Sustainable Use License (fair-code) | 4.7/5 (312) | Operated |
| Supabase | You need the Clay table: one base every workflow reads and writes | Free plan, 500 MB | Open source (Apache-2.0) | No review page yet | Operated |
| Playwright | Your research step keeps hitting empty pages | Free, runs on your machine | Open source (Apache-2.0) | No review page yet | Operated |
For open-source projects, the price column is what running them costs us. Prices and ratings come from each tool's own review page. Market rating is the public review average; the Enjyn Score is our own grading.
n8n
Free, then €20/mo · Enjyn Score 9.4/10
This is where the Clay columns go. Our nightly n8n flow sources leads, tries Apollo, then Findymail, then Prospeo, so one provider missing a contact does not kill the row, and it dedupes against the CRM and the sequencer before anything is paid for. Execution-based pricing is what makes it work: a long workflow costs the same as a short one, the opposite of per-action credits. Self-hosted, the code costs nothing.
Watch out. Fair-code, not open source in the strict sense: the Sustainable Use License lets you run it for your own business, not resell it as a service. Self-hosting also makes the upgrades, the backups and the cron that silently stops firing your problem; we run a watchdog outside n8n for exactly that.
Full n8n review →Supabase
Open source · Apache-2.0
A Clay table is really a database with a nice grid on top. Ours is Supabase: one Postgres base of every company and lead we have already sourced, shared by the whole team since October. Every new list is checked against it by normalised email before a single enrichment credit is spent, which is the saving Clay users chase with lookups across tables.
Watch out. No grid, no point-and-click enrichment: you query it in SQL or through your workflows. And size it before you import: after loading our archive we stood at 448 MB of the 500 MB free plan, about 6.5 KB per lead, so a large archive needs the paid plan or a self-hosted instance from day one.
Supabase on GitHub →Playwright
Open source · Apache-2.0
Half the job of Claygent is reading a prospect's website. A plain HTTP request returns an empty shell on sites built in JavaScript, and that is where research quietly fails. Playwright drives a real browser, renders the page and hands back the text. Our personalisation script falls back to it when the plain read comes back empty.
Watch out. A headless browser is slow and heavy compared with a request. Use it as the fallback, not the default: we only render a page after the cheap read has come back empty.
Playwright on GitHub →Our take. Teams leave Clay for the bill, and the bill is mostly data. An AI column on 10,000 rows can drain a month of credits in one run, but the email finders behind the waterfall cost money in any stack. What open source removes is the markup on the plumbing: the table, the routing, the retries, the dedupe. That is real money once you run the same workflow every night. The one piece we did not replace with open source is the AI column: we write research and personalisation lines in a Claude session, from text the scraper brings back. If you will not own the workflows, keep Clay. This stack only pays when someone on the team is happy to read an n8n execution log at 9am.
Frequently asked questions
Is there an open-source version of Clay?
Not one project that does what Clay does. The open-source route is a stack: a workflow tool such as n8n for the waterfall, a Postgres base such as Supabase for the table, and a headless browser such as Playwright for website research.
Does an open-source stack make enrichment free?
No. It removes the cost of the table and the orchestration, not the data. Email finders and databases (Apollo, Findymail, Prospeo in our waterfall) are still paid per lookup. What you stop paying is a credit on every step a workflow takes.
Is n8n open source?
It is fair-code. The code is public and you can self-host it for free, but the Sustainable Use License limits commercial resale, so it is not open source in the OSI sense. For running your own GTM workflows, that makes no difference.
What replaces Claygent?
Two pieces: something that reads the website, and a model that writes from what it read. Playwright covers the first when a site is built in JavaScript. For the second we use Claude in a working session rather than an open-source model, because the lines go out under our name and get reviewed before they do.
When should I stay on Clay?
When nobody on the team will own the workflows. Clay's value is that a GTM person can build a waterfall without writing code. An n8n and Supabase stack is cheaper per run, but someone has to maintain it, fix it when a provider changes its API, and watch the nightly runs.
Not sure what should replace it
We run outbound stacks for B2B teams and pick tools by what they cost per result, not by the pitch.
Book a GTM audit