Edgee Claude Code Compressor V2 - Fewer tokens, same context, 50% cost reduction
by•
Compression V2 cuts coding-agent token bills with three techniques across two layers: sharper tool result trimming, new task-aware tool surface reduction, and output brevity. Drop-in for Claude Code, Codex, OpenCode, and Cursor. Semantically lossless.
Replies
Nice to see actual sign tests and confidence intervals in the writeup instead of a bare headline number. Lossless trimming? Very interesting - I need to check this out!
Edgee
@marcin_dancewicz Thank you very much. Indeed, it seemed very important to us to be precise and thorough in our explanations and benchmarks. We too often see astonishing saving figures thrown around, without any proof or scientific documentation to back them up.
That's clever. Does it figure out the active task on its own, or do you need to pass in context?
Edgee
@dhiraj_patel5 Everything happens automatically, no need to specify anything. Edgee receives the full request from the agent, processes it as a whole, then forwards it to the model. We try to make the experience as smooth as possible.
If you’re using coding agents all day, them tokens add up really fast and cutting them would save a lot of money. Does the token compression ever affect the quality of the responses?
Edgee
Hey @reda_roqai_chaoui . Our token compression layers are made to be as lossless as possible. Feel free to read @0kham blog post who made a lot of benchmark with SWE to evaluate our compression strategies: https://www.edgee.ai/blog/posts/introducing-compressor-v2-three-compression-layers-measured-end-to-end-for-a-50-cost-reduction
GPTEverywhere
V2 stacking tool-result trimming with the other layers for real Claude Code sessions is the kind of boring infra win that actually shows up in the bill. Curious how it behaves on long MCP-heavy runs.
Edgee
@ducan Happy to help you explore our TSR feature with long session. Feel free to reach out with feddbacks ;)
the brevity layer raises a question beyond live readability: when an agent does something wrong three sessions later and you're doing a postmortem, do you still have the verbose reasoning trace anywhere, or did it only ever exist on the wire in brief form. cost optimization and debuggability usually pull in opposite directions - would love a mode that keeps the full narration in a log for audit while only sending the brief version to the model, so you're not trading "cheaper now" for "unexplainable failure later"
50% cost reduction is a concrete promise. Curious how you keep the compressed context from dropping those tiny repo-specific constraints that usually matter most in Claude Code sessions?
Macaly
50% token savings is a real no brainer for anyone runnign coding agents 🔥 congrats!