Stop paying for words your AI never needed.

Every request you send is billed in full — even the parts the answer never uses. CompresLM removes them on the way through. Same answers, smaller bill, one line to connect.

Start free — 1M tokens See the measurements
−98% prompt, best case on repetitive logs
−52% reply length, optional mode
You save
$37.50/mo
$150.00 → $112.50 per month
Model

The slider starts at a measured average. Repetitive text goes far higher; a page of distinct facts goes lower.

The idea in one picture

Compression follows your question

Like a highlighter on a thick binder: only what your question needs goes through — and only that is billed.

The same effect, measured on real documents:

Pointed"When does failover trigger?"

−59%

Topic"How are malformed events handled?"

−39%

Everything"Explain the whole pipeline."

−18%

A log repeating one error 300×"What happened, and how was it fixed?"

−98%
Sent to the model Removed — never billed

In every measured run, the one line that mattered survived. See how it was measured →

Where your prompt goes — and what we write down

One stateless container. Your prompt is trimmed in memory on ordinary CPU, forwarded to your provider with your key, and freed the moment the call returns. The reply passes straight back to you. The only thing written down is a pair of token counts.

UML deployment diagram. Your infrastructure holds your app and your provider key. A single stateless CompresLM container runs four components in sequence: API and auth, the USC engine on CPU where every span is embedded locally with no GPU and no outside call, egress which adds your key and drops it, and a token counter recording tokens before and after. Request memory is freed when the call returns, so the prompt exists nowhere else. A note below the container states that there is no response cache: the reply is handed back and forgotten, never written down, not even for a second. The one store outside the container is Postgres, holding accounts and token counters, with a constraint note that no prompt text, no reply text and no provider key ever reaches the database. The reply travels back to your app byte-for-byte.
Compression happens inside one process, on CPU — the prompt is never handed to another service and never lands in a database. The reply isn't cached either: it goes straight back to you and is gone.

Those two counts are what your savings — and our bill — are computed from. Privacy Policy → · How it works in detail →

One line to install

Works exactly like calling OpenAI, Claude, or a local LLM directly.

# Python — OpenAI SDK
client = OpenAI(
    base_url="https://compreslm.com/v1",
    api_key="YOUR_COMPRESLM_TOKEN",
    default_headers={"X-LLM-API-Key": "YOUR_PROVIDER_KEY"},
)

We never store your provider key. It rides with each request and is discarded after the call.

Or let your coding agent do it

Copy this into Claude Code, Cursor, or whatever you already have open. It walks your agent through the wiring, stops to ask you for the two things it cannot do itself, and finishes by running the same question with and without compression so you see the saving and the answer side by side — on your own material, not ours.

Setup prompt 


    


    

      
Pricing that can only save you money

      

        The first 1,000,000 tokens (≈750,000 words) are free,
        no card required. After that you pay 20% of the money we
        actually save you — a bill that can only exist if the saving
        does.
      

      

        

          1Mtokens
          Free to start, no card
        

        

          20%
          Of dollars saved, after that
        

        

          $0.50/ 1M
          Tokens saved, local models
        

      

      
How the bill is calculated →

    


    

      
What you get

      
Six things that start the moment you change the base URL.


      

        

          −98%
          
The biggest prompts shrink most

          
Logs, tool output, boilerplate — the more it repeats, the less is sent.

        


        

          0new words
          
Nothing is invented

          
We leave words out. We never write new ones in.

        


        

          Betterin some cases
          
Sometimes the answer improves

          
In blind duels the shorter prompt beat the full one — less noise to distract the model.

        


        

          ∞context
          
Send the whole archive

          
Pay only for the part your question actually needs.

        


        

          −52% reply
          
Replies shrink too

          
And output costs several times input — reasoning models bill their thinking as output.

        


        

          0stored
          
Your text stays yours

          
No database, no cache, no training, no third parties.

        

      


      

        

          
Need it fitted to your setup?

          
Closed network, unusual documents, a hand getting it running — tell us and we will help.

        

        Tell us what you need
      

    


    


  
Talk to us

  

    

      
    
        
  • Connect it — we'll get it running against your traffic and check the numbers with you.
    • 
        
  • Fit it to you — closed network, unusual documents, a shape we don't handle yet.
    • 
        
  • Build it — no product to connect yet? Our partners build it, we make it cheap to run.
    • 
      

      
Usually a reply the same working day.

    

    

      

        
        
      

      

        
        
      

      
      

    

  





    

      
Start saving today

      
First million tokens free. No card, no sales call.

      Create your account