Blog
Practical guides on consolidating AI tools, automating workflows, and getting more from your team's AI investment.
168 Integrations, One Plugin at a Time
Four years of building workflow integrations across communication, productivity, CRM, AI, and 160 other services.
5,876 Commits Across Three AI Products
A four-year retrospective on building workflow automation, a multi-model AI dashboard, and an AI agent platform.
Building Custom GPTs Before OpenAI Did
We shipped a custom chatbot builder with RAG knowledge bases in January 2024, weeks before GPTs launched.
Forty Agents and Counting
Building 40+ specialized AI agents with distinct capabilities, from code review to art generation to project management.
From News Scraper to AI Dashboard in 90 Days
The story of how a news aggregation service pivoted to become a multi-model AI productivity platform.
From Subscriptions to Credits
Three billing system rewrites in three years, from direct Stripe to external API to pay-as-you-go credits.
Magic Chat to Unified Threads
How we merged three separate chat systems into one unified threading architecture over eight months.
Pro Search: Teaching AI to Research
Building a search system that combines Google Custom Search with AI summarization and proper citations.
Research Agents Before Agents Were Cool
We built autonomous research agents with Google Search and task planning in May 2023, before the agentic AI wave.
Testing AI with AI
How we built an evaluation framework that uses LLM-as-judge to score our agents and catch quality regressions.
The Deploy Pipeline
18 GitHub Actions workflows, three environments per product, Pulumi IaC, and two releases per day.
The Enterprise Question
When individual users became teams, and teams became organizations, and organizations needed corporate subscriptions.
The Fork That Started Everything
How forking an open-source Zapier alternative in October 2021 became the foundation for a four-year AI platform.
The Infrastructure Nobody Sees
Auth services, message queues, three deployment regions, and the shared layer that keeps three products running.
The Knowledge Base We Rebuilt Twice
From Pinecone vectors to SmartFAQ proxy to VFS with OCR -- three years of evolving our RAG system.
The LikeClaw Variant
How we turned a cloud-dependent AI platform into a self-contained product with local auth and billing in four days.
The Rewrite Decision
Why we chose to rebuild our AI dashboard from scratch in November 2025 instead of iterating on v1.
The Wizard System
How we added human-in-the-loop to workflow automation with five types of interactive wizards.
Thirty Models, One Interface
How we went from GPT-3.5 to 30+ AI models across four providers without building 30 different integrations.
Three Databases, Three Problems
PostgreSQL for workflows, MongoDB for AI, Redis for speed -- how three databases serve three products.
When Autopilot Learned to Think
The story of adding 40+ AI models to a workflow automation platform that started with zero AI capabilities.
