Inspiration AI agents are becoming essential tools customer support bots, code reviewers, content generators, data analyzers. But they all share one problem: runaway costs. A single misconfigured agent can burn through hundreds of dollars in minutes. There was no simple way to monitor and control spending across multiple agents and models in one place. We wanted to build something that gives developers and teams peace of mind set a budget, let the agent run, and know it will stop before it drains your wallet. What it does AgentWatch is a real-time dashboard for monitoring and controlling AI agent spending. You can track costs across all your agents (GPT-4, Claude, Gemini, Llama), set daily budget limits with configurable alert thresholds, and auto-stop agents when they exceed their budget. It shows cost trends with 7D/30D/90D charts, sends alerts when agents approach spending limits, and logs all activity in a live feed. One dashboard to see everything, what's running, what's costing money, and what got stopped. How we built it Built with vanilla HTML, CSS, and JavaScript no frameworks, no build tools, no dependencies. Runs entirely in the browser with local storage for persistence. Chart.js for the cost visualizations. Hosted on GitHub Pages so anyone can access it instantly. Integrated Novus.ai for product analytics to understand how people use the dashboard. Challenges we ran into Making the real-time cost simulation feel realistic without a backend. Designing auto-stop guardrails that trigger at exactly 100% budget without blocking legitimate usage in progress. Getting the theme toggle to work smoothly across all components. Making the dashboard responsive so it works on different screen sizes. Accomplishments that we're proud of A fully functional cost monitoring dashboard built from scratch with zero frameworks. The auto-stop feature that actually prevents budget overruns. The clean dark/light theme. And the fact that it's live and working right now not a mockup, not a wireframe, a real product you can use today. What we learned That you don't need a complex tech stack to build something useful. Vanilla JS is still powerful. Simple, focused tools solve real problems. And budget guardrails for AI agents is a problem people actually care about. What's next for AgentWatch Backend integration for real API cost tracking (OpenAI, Anthropic, Google). Multi-user support with team dashboards. Historical cost reports and analytics. Slack/Discord notifications for budget alerts. And an API so other tools can integrate cost guardrails into their workflows.

Built With

Share this project:

Updates