CodeBurn is an interactive terminal UI (TUI) dashboard that gives you full visibility into where your AI coding budget is going. It parses the session logs from Claude Code, Codex, and Cursor, then surfaces cost breakdowns by project, model, session, and time period — right in your terminal with a polished, color-coded interface. Whether you're a solo developer burning through a monthly API quota or a team running 15+ projects across multiple models, CodeBurn makes cost observability a first-class citizen in your workflow.

• Multi-agent support out of the box — Seamlessly tracks Claude Code, OpenAI Codex, and Cursor IDE sessions. Run it once and it auto-discovers all your project logs, no manual configuration required.
• Per-project drill-down with --project and --exclude — Focus on a specific workspace or exclude noisy test repos. Shipped in v0.6.1 (PR #52 by community contributor @mallek), addressing the #1 feature request from power users running 170+ sessions/month.
• Cost-per-session metric and top expensive sessions panel — The avg/s column and Top Sessions panel let you immediately spot runaway sessions, identify which projects are budget sinks, and make data-driven decisions about which models to use where.

A user running 15+ projects with 170+ Claude Code + Sonnet sessions/month opened the ticket with six concrete feature requests. The maintainer shipped four of them in a single release cycle:

@sterngold (2026-04-16): "First — great tool. Installed via npx, zero friction, immediately useful on a 15-project setup running 170+ sessions/month across Opus and Sonnet."

@AgentSeal (2026-04-17): "Hi @sterngold, circling back with status. Four of your six proposals shipped in 0.6.0 / 0.6.1. ... Looking forward to date ranges and model comparison when they land. Great project."

@sterngold (2026-04-18): "Four out of six in a single release cycle is genuinely impressive. Per-project drill-down and cost-per-session are exactly what I needed most."

Translation: A power user hits a real pain point, the community and maintainer collaborate, and four features land in days. That's what open-source velocity looks like.

@billyg555 (2026-04-19): "Could you consider adding a Windows widget via a Rainmeter skin, similar to the macOS menubar integration? CodeBurn already has CLI/JSON output, so a widget would just need to poll the output at intervals."

@AgentSeal (2026-04-20): "This is on the roadmap. We'll track it as a Windows-specific enhancement in the next milestone. In the meantime, you can use the codeburn status --json output piped to any dashboard tool."

CodeBurn fills a critical gap in the AI coding toolchain: nobody else was giving developers real-time, per-project cost visibility across multiple AI coding agents. With its zero-config TUI, cross-agent support, and active community-driven feature pipeline, it's becoming the de facto cost observability layer for Claude Code and Cursor power users. The maintainer is responsive, the roadmap is public, and contributions are welcomed — if you're serious about AI-assisted development at scale, this is worth a look.

Try it: npx codeburn today

6,233 stars · @getagentseal · https://github.com/getagentseal/codeburn