ModelShortlist is a local read-only MCP context layer, not an inference proxy.
ModelShortlist vs model routers
ModelShortlist recommends. It does not route inference.
Model routers and ModelShortlist can both help with model choice, but they operate at different points in the stack. ModelShortlist gives your host AI current evidence for an explicit workload; it does not sit in the inference path.
A router may dispatch individual requests automatically at runtime.
ModelShortlist is useful when a human or AI wants an explainable shortlist before choosing a model.
The local/BYOK architecture keeps ModelShortlist out of the model traffic path.
01
What ModelShortlist does
The MCP retrieves current model facts, attaches independent Artificial Analysis evidence when identities match confidently, applies explicit hard constraints, and returns enough context for the host AI to explain a shortlist.
No hosted ModelShortlist inference endpoint.
No automatic request-by-request model switching.
No permanent global score that overrides workload context.
02
What a model router typically does
A router usually participates in the runtime path and chooses where a request should be sent based on routing logic, policy, availability, cost, quality, or other signals. That can be valuable when automatic dispatch is the product requirement.
03
When ModelShortlist is the better fit
Use ModelShortlist when you want an evidence-backed recommendation you can inspect before configuring an agent, application, workflow, or provider. The result can inform a router configuration later, but ModelShortlist itself does not become the router.
Try it in your assistant
Turn the concept into a current decision.
These prompts ask for workload-specific reasoning rather than a permanent ranking. The answer can change as benchmark and operational evidence changes.
“I am choosing a default model for this coding agent. Shortlist the best current options and explain the tradeoffs before I configure routing.”
“Compare these OpenRouter models for my workload and tell me which hard constraints eliminate candidates.”
“I already have a router. Which models should be in its candidate pool for this workload, based on current benchmark and operational evidence?”
Keep exploring
Why ModelShortlist exists
Understand the workload-specific recommendation philosophy.
Read moreOpenRouter model comparison
Use current OpenRouter facts to build a shortlist without routing inference.
Read moreModelShortlist vs leaderboards
See the difference between evidence, ranking, and workload judgment.
Read moreAsk the current model market, not yesterday's blog post.
ModelShortlist runs locally, uses your Artificial Analysis and OpenRouter keys, and gives your host AI current evidence for the workload you actually care about.
Open the install configurator