GTM Studio is a Clay/Sculptor-grade enrichment table that runs on your machine. Rows are companies or people. Columns are enrichment steps — tools, AI, waterfalls — that run as a dependency graph. Your providers, your data, your spend.
Operators rebuilt the same enrichment plays column-by-column in expensive cloud tools. GTM Studio is the same shape — but it runs locally, talks to the real providers, and gates spend so you only pay for rows that qualify.
SQLite on your disk is the source of truth. Works fully offline; optionally syncs each client to its own Supabase schema in the background.
86 built-in tools — Blitz, Kitt, LeadMagic, IcyPeas, Serper, Spider, MillionVerifier and more. Add any HTTP API as a custom tool. Your keys, your bill.
A cheap AI or formula check gates each paid column. Rows that don't qualify are skipped before any spend — the gates are the cost model.
A Claude agent grounded in your GTM playbook and aware of the live table. Ask it to build the columns, fix a chain, or run the enrichment for you.
Add rows, then columns enrich them in dependency order. The ICP gate decides who's worth spending on — unqualified rows skip the paid people + email columns entirely.
| CompanyINPUT | DomainINPUT | FirmographicsTOOL · leadmagic | ICP TierAI · gpt-4o-mini | Work EmailWATERFALL |
|---|
Each client is an isolated workspace — its own tables, entities and (optional) cloud schema.
Blank, from a recipe, a CSV import, or a Source (find companies / people from a provider).
Each column declares its inputs. The engine computes the run order — no manual wiring.
A run_condition skips a column for rows that fail it. Cheap filters run first; paid steps stay behind the gate.
Cells fill with a value, a status and a cost. Re-runs skip fresh cells and anything you edited by hand.
Push qualified rows to a Destination — webhook, CRM, or a campaign — or export CSV.
Mix them freely. Every column reads upstream cells; the graph keeps them in order.
Manual data, CSV, or a Source. The seeds everything else reads from.
One provider call — enrich a company, find an email, search the web. 86 built in.
An LLM prompt over upstream cells: classify, score, write a line, extract.
Try providers in order; stop at the first that returns. The email/phone play.
Deterministic JS — build a query, derive a domain, normalize a value. Free.
Built-in cleaners (name, company, email) with an optional LLM fallback.
Read a cached field off the row's company/contact entity. No cost.
Offload a heavy enrichment to a managed cloud job and stream results back.
A waterfall column tries each provider in order and stops the instant one returns a usable value — so you never pay the third finder when the first one had it.
Pre-built column chains — pick one and it instantiates a table ready to run. The flagship GTM pipelines come gated out of the box.
Every edit writes to local SQLite instantly. If a workspace is connected, the active client replicates to its own Postgres schema every 30s — row-level last-write-wins. Pull the plug and nothing breaks.
The Companion is a Claude agent that runs inside your GTM playbook checkout and can drive the live table through the archon CLI — read context, add columns, run, approve gates.
Run the desktop installer. Your data lives at %APPDATA%/archon.
Name your workspace from the welcome screen.
e.g. Pipeline · Company → People — it builds the columns for you.
Drop provider API keys in Settings → API Keys. Only the ones a recipe uses are needed.
Paste, import, or use a Source — then hit ▶ Run Enrichment.
“Build me a list of Series-B fintechs with verified emails.”