Use cases · Agents

Your agent reads 40 pages to answer one question

Every tool call returns everything it found. The next step needs three lines of it — and you pay for all of it, on every hop.

−97% Of an incident log, root cause still found
13.7k→ 0.4k Tokens reaching the model

What you hand over, what comes back


In a chain of agents it compounds

Each hop carries the last one's output forward. By the fourth, most of the prompt is work already done — and billed again.

Researcher
step 1
Analyst
step 2
Reviewer
step 3
Writer
step 4
Prompt as it arrives What the step actually needs

Illustrative shape, not a measurement — how much is removed at each hop depends on your material. The playground gives you the real number.


Connect it

Change the base URL. Keep your SDK, your model, your key.

# 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"},
)

# Tool output goes in `context`, the step stays the message.
resp = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Which service caused the timeouts?"}],
    extra_body={"context": tool_output},
)

Long tool results are the best case there is. See the request shapes for the details.

Run this exact case

A day of incident chatter, one question, both answers side by side.

Try it — no account

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.