Yes — that's exactly why I built it. Log-based tools only account for roughly half of the token cost I wanted to see where the other ~50% goes: system prompts, tool schemas, MCP overhead, plus per-tool stats on tool use and tool output sizes(these tool use do show up in logs, but no one attributes them, and the logs have no exact schemas).
You're right about the tradeoff too. The local proxy + wrappers do add friction, and that's hard to avoid when you sit on the wire. I've tried to minimize it by automating as much of the setup as possible in the installer, and adding a lot of error handling and recovery.
Really glad the distinction landed — this is exactly the framing I'd hoped people would see. Would love to hear any feedback as you use it, and let's stay in touch.
Not really. Codeburn tells you how much a chat/code cost, we tell you what's inside your request from API: how much does system prompts, schemas, MCP blocks, thinking, cache reads/writes, tool calls, and outputs cost.
Yes — that's exactly why I built it. Log-based tools only account for roughly half of the token cost I wanted to see where the other ~50% goes: system prompts, tool schemas, MCP overhead, plus per-tool stats on tool use and tool output sizes(these tool use do show up in logs, but no one attributes them, and the logs have no exact schemas).
You're right about the tradeoff too. The local proxy + wrappers do add friction, and that's hard to avoid when you sit on the wire. I've tried to minimize it by automating as much of the setup as possible in the installer, and adding a lot of error handling and recovery.
Really glad the distinction landed — this is exactly the framing I'd hoped people would see. Would love to hear any feedback as you use it, and let's stay in touch.
A colleague recommended this to me after I complained that Fable 5 usage was getting hard to understand.
I tried it for a few sessions, and the useful part was seeing the cost without reading the logs.
Sharing here in case others using coding agents have run into the same problem.
Just say you made it, that's ok here.
Lol, actually not, it's mine.
Is it another codeburn?
Not really. Codeburn tells you how much a chat/code cost, we tell you what's inside your request from API: how much does system prompts, schemas, MCP blocks, thinking, cache reads/writes, tool calls, and outputs cost.
[flagged]