> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openserv.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Why SERV Reasoning
> Better results from the models you already use.
SERV Reasoning is an inference API for LLMs. Each call runs through a structured reasoning step, following a bounded graph, that improves the consistency and accuracy of outputs.
It sits between your application and the model: the same request returns **more reliable results, at lower cost**, with no change to how you build.
```mermaid theme={null}
flowchart LR
A["Your app"] -->|same request| B["SERV Reasoning
structured reasoning layer"]
B -->|one endpoint| C["Any model
multi-provider catalog"]
```
Every request runs through a reasoning step — more accurate and consistent outputs than calling the model directly.
Pair a smaller model with SERV Reasoning to reach a larger model's quality. Savings scale with volume.
Change the base URL and API key in your existing OpenAI or Anthropic SDK. Nothing else changes.
A broad catalog of models behind a single endpoint and a single API key.
Compare any model with and without SERV Reasoning in the Playground. No code required.
One layer between your app and the model. More reliable, cheaper, zero rewrite.
These are available as [SERV Tools](/serv-reasoning/tools): turn them on or off by adding a tool to any request, in any SDK.
Protect your system prompt from injection-based leakage. Opt in with the `serv_prompt_guard` tool.
On by default: SERV blocks output that would reveal your system prompt. Disable it per request with the `serv_disable_content_filter` tool.
Difficult tasks reach higher accuracy through a validate-and-iterate loop. Opt in with the `serv_shadow_agent` tool.
Full technical diagram of SERV Reasoning engine:
.png?fit=max&auto=format&n=Dt_AFMyZaE1zysRx&q=85&s=208340de675ac90f693724f5b4f5a542)