# sim-lib-agent-runner-http
In one line: The connector that lets SIM reach hosted and local models through provider-shaped HTTP profiles.
## What it gives you
This crate lets SIM talk to a model that lives behind a web address. You point it at a service, and it handles the back-and-forth: sending your request, reading the reply, and passing along partial text as it streams in so you do not wait for the whole answer. It understands hosted OpenAI and Anthropic services, local Ollama, LM Studio, and Lemonade servers, and other OpenAI-compatible endpoints. It also takes care to keep sensitive parts of a request, such as keys and private content, from leaking into logs and traces. From SIM's point of view, each provider becomes just another runner it can call.
## Why you will be glad
- You can use hosted or self-run model services from SIM without writing network code yourself.
- Answers stream in as they are generated, so long replies feel responsive.
- Sensitive request details are kept out of logs, reducing the chance of an accidental leak.
## Where it fits
This is one of SIM's model runners, the kind that reaches a model through HTTP. It follows the shared runner contract, so anything that can call a runner can call a hosted service or loopback model server through this crate. Sibling crates cover models that run in the same process or as a local subprocess; this one covers the providers you reach over HTTP.