Poco AI / benchmarks
We take the padding. Not the signal.
Commands
35.5%
median saving
We drop the command output the conversation never needed.
The numbers →Commands / receipts
Measured with the o200k_base tokenizer, against what the agent would have read without Poco.
← BackFiles
Up to 70%
Documents are converted into AI-friendly formats.
The method →Files / method
PDF, Word, PowerPoint, Excel and HTML all become Markdown before the model reads them.
A PDF is the hungriest of them, and gives back the most: up to 70% fewer tokens, with the content intact.
← BackMCP servers
One command prices every MCP server you have installed, and flags the ones you never call.
The method →MCP servers / method
A server's tool definitions sit in the always-loaded system prompt. You pay for them every turn, called or not.
We price each server, then show you the ones you never call.
← BackPrompt
Your prompt, rewritten to say the same thing in fewer tokens.
The method →Prompt / method
Much of a prompt is written for a reader who isn't there. Throat-clearing, hedging, restated context, politeness: the model pays for all of it and needs none of it.
We cut the scaffolding and keep the instruction.
← BackA median of 35.5% across real commands.
Up to 70% fewer tokens. PDF, Word, PowerPoint, Excel, and HTML files are converted to Markdown before the model reads them, and PDFs see the largest reduction.
Yes. A server's tool definitions sit in the always-loaded system prompt, so you pay for them every turn whether you call them or not. Poco prices every installed MCP server and flags the ones you never call.
Yes, in beta. Poco rewrites your prompt to say the same thing in fewer tokens - cutting scaffolding like restated context, hedging, and politeness while keeping the instruction. It doesn't have a measured saving in Claude Code yet, which is why no percentage is quoted for it.