Beton

2 years ago I started writing on open-source pricing on my personal Substack read through pricing pages & licenses to find out if the paid tier is worth it or if I use self-hosted version if I need it someday I just repeated the research for 20 OSS companies for 2026, and pulled the 2023 numbers off the Wayback Machine to see what actually changed what I didn't expect: - where metering matured, prices went down. PostHog's per-event price fell 36%, ClickHouse storage 28%. - "AI credits" is now a standard line item. It didn't exist in 2023. - SSO tax is real and creeping. SSO costs $9/seat at Twenty and $25,000/year at Grafana Labs; Cal.com quietly moved it from its $12 tier to its $28 one. all the content is already open: the full write-up with charts, all 20 teardowns, and the raw dataset on github comment "open source pricing" if you want to support us, link's in the comments👇 #opensource #pricing #saas #revops #buildinpublic #calcom #metabase #grafana #twentycrm #clickhouse #posthog #tooljet #coolify #appsmith #novu #windmill #chatwoot #dify #langfuse #infisical #firecrawl #plane #n8n #

just cut Beton's agent token consumption 4x woke up to a slack message from our CTO Alexandr Markelov Alex managed to keep the quality of analytics our agent does by making it waste less so, we (and Beton's self-hosters) can use more expensive LLMs in production, which grows the overall product quality with a blink of an eye proud of the team!

Beton has quietly shipped some things in the past month slack integration, mcp server, data bugs moved our blog off substack in a desperate move to fix SEO issues with main website. having fun with astro.js and Resend coming soon: postgres + hubspot as a data source, hubspot as a signal destination, multi-user workspaces + stat tools for signal backtesting cherries on the cake: we're open-sourcing: - agentic backend behind Beton (Google ADK, AGPLv3). it's the core of our revenue intelligence product - library for generating synthetic product events (Python, MIT). it's a internal tool that Alexandr Markelov built while he waited for me to get real customer data both already live on (but thinking of a better name than inspector-ml-backend) full update: https://lnkd.in/ejFR-6aP

we deleted the login screen feels like a fringe idea, but it can work always wanted let people see the product before asking for their email/phone unfortunately, there will be no instant time-to-value with us – calculations and research take time even for agentic systems now you the whole Beton's app is visible. see before you even sign up real UI, mock data, register at any time with Google after signup, a 6-step wizard that actually explains what it wants from you including a live preview of how your deals will look in your CRM when Beton creates them (see below) link to the newsletter in comments, ofc

new onboarding, MCP, Firecrawl integration – one more Beton newsletter after we launched our new app, we found that signups could not get through the onboarding app was mostly empty screens + some flow that asked for API keys to your analytics I spent a week in front of 3 Claude Codes to fix this: https://lnkd.in/dtVPZHAG #revops #revenueops #salesops #expansion #upsell #reactivation

cooking a big release for Beton, sit tight

most open source companies pick one license and hope for the best (not counting enterprise ones) Firecrawl runs two — and that seems to me like a great strategy AGPL on the core scraping engine. can't fork it into your product without open sourcing everything MIT on the enrichment layer. fork it freely — but every query pays Firecrawl for scraping and OpenAI for tokens one license is a wall. the other is a funnel. together, it's one of the cleanest GTM plays in open source infra full pricing teardown in comments

sales teams try to qualify leads based on product usage at least from what I've seen ofc guess which events matter, assign weights, hope it works it doesn't signals are buried in behavioral combinations no human would think to look for but they can predict buying intent Beton crawls your PostHog data, learns your data model, and discovers patterns that correlate with plan upgrades or closed/wons — automatically once we spot them again, we create a deal in your CRM (Attio for now, my DMs are open for requests) no spreadsheet scoring, no manual rules - the signals find themselves PS we're open source → getbeton.ai