WEBVTT

1
00:00:00.259 --> 00:00:03.728
Most of what you ask an AI doesn't need the most expensive model.

2
00:00:03.728 --> 00:00:04.528
SemanticRouter.

3
00:00:04.528 --> 00:00:08.797
NET looks at each prompt, and sends it to the cheapest model that can handle it.

4
00:00:08.797 --> 00:00:12.479
That might be a local model, your Claude subscription, or a paid API.

5
00:00:13.386 --> 00:00:15.327
Getting set up takes about a minute.

6
00:00:15.327 --> 00:00:16.406
First, install the .

7
00:00:16.406 --> 00:00:16.891
NET tool.

8
00:00:17.794 --> 00:00:19.435
Then run the init command.

9
00:00:19.435 --> 00:00:22.148
It looks at what's already on your machine.

10
00:00:22.148 --> 00:00:28.459
Here, it finds LM Studio with two local models, and the Claude Code CLI you're already signed in to.

11
00:00:28.459 --> 00:00:31.930
Then it writes a config that puts each model on a tier.

12
00:00:31.929 --> 00:00:33.102
Now start the router.

13
00:00:33.102 --> 00:00:38.073
It speaks both the OpenAI and the Anthropic API, and it tells you exactly how to connect.

14
00:00:38.578 --> 00:00:43.773
In a second terminal, point Claude Code at the router, and pick semantic router as the model.

15
00:00:43.773 --> 00:00:45.728
Claude Code works just like before.

16
00:00:45.728 --> 00:00:46.845
The only difference?

17
00:00:46.845 --> 00:00:49.582
The router decides which model answers each turn.

18
00:00:50.807 --> 00:00:53.789
Ask something simple, and a small local model answers.

19
00:00:53.789 --> 00:00:56.330
That's free, and it never leaves your machine.

20
00:00:56.932 --> 00:00:58.017
A quick coding task?

21
00:00:58.017 --> 00:00:59.591
That goes to the local coder.

22
00:01:04.993 --> 00:01:10.311
But a tricky debugging question goes to Claude, where your subscription is actually worth spending.

23
00:01:15.730 --> 00:01:18.030
And when you hit your Claude session limit?

24
00:01:18.030 --> 00:01:21.934
The router cools Claude down, and keeps you working on your local models.

25
00:01:21.934 --> 00:01:23.271
You don't change a thing.

26
00:01:25.307 --> 00:01:28.293
The dashboard shows every decision as it happens:

27
00:01:28.293 --> 00:01:33.474
which route matched, which model answered, and how much of it stayed on your machine.

28
00:01:33.474 --> 00:01:35.302
And it's not just Claude Code.

29
00:01:35.302 --> 00:01:38.776
Any OpenAI or Anthropic client can use the same endpoint.

30
00:01:39.379 --> 00:01:40.924
That's SemanticRouter.

31
00:01:40.924 --> 00:01:41.204
NET.

32
00:01:41.204 --> 00:01:43.592
It's open source and MIT licensed.

33
00:01:43.592 --> 00:01:49.138
Install it with one command, and stop paying frontier prices for small prompts.