Beton detects revenue signals in your product data automatically — no rule-writing, no playbook configuration — and routes them to your CRM. Here's what's under the hood.
The agent reads your schema, proposes behavioral hypotheses, and backtests each one before it ever fires.
Beton's agent proposes signal hypotheses tailored to YOUR schema. You review and approve — no rule-writing, no analytics SQL.
The agent reads your PostHog event taxonomy or Postgres schema and proposes testable behavioral hypotheses — phrased in plain English, scoped to your product. You review the candidates, approve the ones worth tracking, and skip the rest. No SQL, no rule files, no whiteboarding sessions to figure out which events matter.
Every signal hypothesis is scored on your last 90 days of events — precision, recall, lift over baseline — before it routes anything to sales.
The agent proposes hypotheses, then validates each one against your own historical data. You see precision, recall, and lift over baseline before promotion. Hypotheses that don’t clear your bar never become signals — and the ones that do come with the math attached.
When a signal fires, Beton pushes it where revenue work already happens — CRM records, webhooks, your stack.
One detected signal fans out to Attio, Apollo, and any HTTPS endpoint via Webhooks. HMAC-signed, retried, and logged.
A single behavioral signal triggers updates across every connected destination — Attio for CRM context, Apollo for contact enrichment and sequence enrolment, plus any HTTPS endpoint (n8n, Zapier, Make, your own API) via the Webhooks integration. HMAC-signed payloads, retries with backoff, full delivery logs.
Upsert Companies, People, and deal records in Attio when buying signals fire. Append-only on existing fields — never overwrites user-entered data.
When Beton detects a revenue signal, it upserts the matching record in Attio by domain or email and appends the signal context as custom attributes. Sales reps see what triggered the signal directly on the record — no tab switching, no manual lookups, no overwritten fields.
Connect the data sources Beton reads from and the destinations it routes to. No ETL to maintain.
Connect PostHog directly, or any Postgres warehouse with a read-only role. Same hypothesis-and-backtest flow on either source.
Beton reads from PostHog (Cloud or self-hosted) via API, or from any Postgres database (replicas, Supabase, RDS, Cloud SQL) via a read-only role. Same agent, same flow on both — no data export, no migration, no extra warehouse load.
Open-source, self-hostable, and built to run on data you already own.
AGPLv3 source on GitHub. Self-host on your own infra — no data leaves your stack — or run on our cloud. Same code either way.
Yes, Beton is a scoring vendor — but a self-hostable one. Run the whole agent on your own VM under AGPLv3, with no data egress, no vendor lock-in, and the same code as our managed cloud. Security review comes free.
Connect PostHog or Postgres and the agent proposes its first signals in minutes. Free to self-host.