platoseedplatoseed
LLM context compression for better accuracy
Compresr provides an API that compresses LLM context without losing what matters. It’s a drop-in for agents and RAG that cuts token costs and improves accuracy.
Compresr offers context compression for LLMs to reduce prompt tokens while preserving or improving accuracy. It provides both hosted API access and on-prem deployment, with transparent per-million-token pricing and credits on signup.
Compresr compresses the input context before querying the LLM, returning only tokens that meaningfully influence the answer. It supports loading a document (PDF/MD/TXT) and a question, then provides a cost and token receipt, runs the compressed context through the LLM, and returns the answer. It can be used via a hosted API, drop-in or SDKs (TypeScript/Python), or deployed on-prem in a VPC with dedicated throughput SLAs. It offers per-million-token pricing, a $10 free credit on signup, and on-prem/private deployment options with tailored support.
Who it’s for: AI teams and enterprises using large language models who want to reduce token costs and improve efficiency without sacrificing answer quality, including regulated industries requiring on-prem deployment.
Backed by Y Combinator; mentions of on-prem deployments, pricing, live demos, and sign-up free credits indicate early commercial traction and growing product familiarity.
CEO @ Compresr. Previously researched LLM context compression as an EPFL PhD (Switzerland). Former Microsoft and Philips Research.
CAIO @ compresr.ai. On a mission to make every token count | EPFL CS | prev UBS
Co-founder & COO @ Compresr (YC W26) | EPFL Data Science Masters | ex-Bell Labs
Cofounder and CTO @ Compresr. Previously worked in research at EPFL’s DLab and AXA, focusing on efficient ML systems and prompt compression.
Cut token spend, reduce latency, boost generation quality.
Compresr offers an API and an open-source compression proxy to reduce long-context in LLM pipelines, enabling 100x context compression to improve generation speed and cost. It targets Claude Code, Codex, and OpenClaw users and promises seamless integration with agentic workflows.
Formerly “Cmprsr” · why startups rename →
Compression middleware that improves LLM outputs
Systems log compression for agents.
