Quick Overview
In this build review and demonstration, creator Nate Herk presents SnagTime, a custom open-source scheduling application built using AI coding agents. The video outlines how the application replicates commercial booking platforms, details the agent prompt architecture, and reviews the compute metrics and token costs of the project.
Key Points
- 1.Using AI coding agents, a full-featured scheduling application named SnagTime was built to replicate core Calendly and Cal.com workflows.
- 2.The application supports custom branding, variable duration booking slots, Google Calendar two-way synchronization, and paid bookings via Stripe.
- 3.The development process ran through a structured four-phase loop consisting of research, planning, building, and automated agentic testing.
- 4.The build required 334 aggregate hours of autonomous agent execution across 76 sub-agents over 5 days and 5 hours of actual runtime.
- 5.Although the computed token cost totaled $14,982.96, it was executed primarily under a $200 per month Codex subscription with minor usage credit overages.
- 6.Autonomous agents excelled at finding functional software bugs but required manual human intervention to address user experience design and interface latency.
Summary
The video demonstrates building a complete, self-hostable clone of Calendly named SnagTime using autonomous AI coding agents. The project originated as an experiment to replicate high-value scheduling software without ongoing SaaS subscription fees, given that commercial competitors like Calendly and Cal.com hold high valuations and require paid subscriptions for advanced features.
Demonstrating the completed application running locally, the walkthrough shows user registration, workspace branding customization, team member management, and event type configuration. Event types support free bookings as well as paid appointments powered by Stripe checkout. A live booking demonstration illustrates smooth two-way synchronization with Google Calendar, showing booked slots disappearing in real time and calendar invites arriving directly in Gmail.
The development process was initiated through a goal prompt dividing the build into four distinct phases: research, planning, building, and testing. During research and planning, the agents analyzed user feedback on existing scheduling tools and outlined system architecture, integrations, and database schemas. During the build phase, dozens of sub-agents operated in an autonomous testing loop, simulating admin setup and booking flows, identifying bugs, and patching the codebase without human intervention.
Despite the agents successfully diagnosing functional errors, human intervention was required to improve user experience and performance. The app was rebranded from its original agent-assigned name, TempoCove, to SnagTime. When the interface exhibited input lag and sluggish response times, targeted prompts were dispatched to reduce load times to milliseconds. The creator also redesigned the booking flow navigation after realizing the agents had implemented unclickable progress indicators that hindered natural user interaction.
- 1.Initial Goal Prompt. Initiated the four-phase automated cycle covering competitor research, system planning, Next.js application coding, and swarm-based testing.
- 2.Rebranding and Redesign. Renamed the software from TempoCove to SnagTime and overhauled visual aesthetics and dashboard layout.
- 3.Latency Optimization. Resolved interface lag on booking pages, forcing sub-millisecond execution and smooth input responsiveness.
- 4.User Experience Refinements. Reconfigured step-by-step navigation in the booking funnel so header progress indicators functioned as intuitive clickable back buttons.
In total, the build required 5 days and 5 hours of actual runtime, aggregating 334 compute hours across 76 unique sub-agents. The project processed 32,103,178,736 input tokens and 47,364,103 output tokens, equaling $14,982.96 in nominal API costs. Because the workload ran under a flat $200 per month Codex subscription, the actual cash cost was confined to the subscription fee plus approximately $150 in additional usage credits.
Features and Core Functionality
SnagTime replicates essential scheduling software features, including multi-duration booking links, Google Calendar synchronization to prevent double-booking, and Stripe integration for paid consultations. The administrative dashboard allows organizers to configure workspace branding, invite team members with distinct permission levels, set custom weekly working hours, apply date overrides, and add buffer times between meetings.
The Agentic Development Loop
Development began with an initial prompt instructing AI agents to execute a four-phase workflow: research, planning, building, and testing. The agents autonomously researched competitor pros and cons, planned database and integration architectures, wrote application code, and executed continuous testing loops where dozens of sub-agents simulated user signups and booking flows to identify and fix software bugs.
Human Refinement and UI Optimization
Human oversight was necessary to resolve flaws that automated testing overlooked. The creator rebranded the app from its initial name, TempoCove, to SnagTime and adjusted the user interface after noticing unnatural navigation patterns, such as unclickable progress bar steps. Additional prompts were issued to resolve interface lag and improve booking page responsiveness down to millisecond load times.
Development Stats and Inference Costs
Building SnagTime consumed 32,103,178,736 input tokens and 47,364,103 output tokens across 76 unique sub-agents, accumulating 334 hours of parallel compute time. At standard API rates, the inference amounted to $14,982.96, but it was accomplished under a $200 monthly Codex subscription plan with roughly $150 in out-of-pocket usage credits.
The Bottom Line
The video establishes that autonomous AI agents can construct fully functional, production-ready SaaS alternatives from scratch with minimal manual coding. It demonstrates that while agents handle backend architecture and automated regression testing effectively, human oversight remains vital for usability polish and design sensibility. The project leaves open the long-term maintenance overhead and bug remediation required when scaling such self-hosted applications to enterprise workloads.
FAQ
What is SnagTime and how was it built as a clone of Calendly?
SnagTime is an open-source scheduling web application built using AI coding agents in Codex that replicates the core features of Calendly, including customizable booking links, Stripe payments, and Google Calendar synchronization.
How much did the AI inference cost to build the SnagTime application?
The nominal API inference cost was $14,982.96 across over 32 billion input tokens and 47 million output tokens, but it was executed under a $200 per month Codex plan with approximately $150 in extra usage fees.
How long did it take AI agents to complete the SnagTime build process?
The agents ran for 5 days and 5 hours of active execution time, accumulating 334 aggregate hours across 76 sub-agents working in parallel.
What four phases did the AI agents follow during the initial goal prompt?
The initial prompt instructed the agents to execute four structured phases: research, planning, building, and autonomous testing loops.
Why was human intervention necessary despite the automated agent testing loop?
Human intervention was required to fix UI lag, rebrand the application from TempoCove to SnagTime, and make progress bar elements clickable for an intuitive user experience.
Worth watching for
Developers, technical founders, and automation engineers interested in using autonomous AI coding agents to build and self-host complex web applications.
- ai-coding
- codex
- calendly-clone
- software-development
- automation
- agentic-workflows