Unlimited Claude Code For Free (Runs On Your Computer) — Transcript

Full transcript

1. 0:00Claude Code is the most powerful coding
2. 0:02agent in the world. The problem is it
3. 0:05costs a fortune. You pay every month and
4. 0:09the second you hit your limit mid-task,
5. 0:11it just stops. You can either pay more
6. 0:14or wait. So, here's the good news. You
7. 0:17can now run Claude Code completely free
8. 0:21on a model that lives on your own
9. 0:23machine. No, this is not a Claude Code
10. 0:25knockoff. This is Claude Code but on a
11. 0:28different AI model. So, in this video,
12. 0:31I'm going to install [music] it, set it
13. 0:33up, and show you exactly how to run it.
14. 0:36By the end, you'll have the same Claude
15. 0:39Code you already use except [music]
16. 0:41it will cost you nothing. So, let's get
17. 0:43started. Here's how this works. Claude
18. 0:46Code is really two separate things in
19. 0:48one. There's the AI model, aka the brain
20. 0:52that does the thinking, and Claude Code
21. 0:54that's Fable, Opus, Sonnet. That's the
22. 0:57Anthropic model family. And then,
23. 1:00there's the tool wrapped around it,
24. 1:02Claude Code. Some people call it the
25. 1:05agent harness, which is just a fancy way
26. 1:07of calling the app that wraps around the
27. 1:09AI model. When you pay Anthropic, you're
28. 1:12paying for the brain, not the tool. The
29. 1:15tool runs on whatever brain you point it
30. 1:18at. So, all we do is keep Claude Code
31. 1:21exactly as it is and point it at a brain
32. 1:24that runs on your own computer instead,
33. 1:27for free. And the thing that makes this
34. 1:30possible is Ollama. It runs AI models
35. 1:33locally, right on your machine, and
36. 1:36recently it started speaking Claude
37. 1:38Code's own language. So, the two connect
38. 1:41directly with nothing in between them.
39. 1:43So, let me show you how to build it.
40. 1:46First, you install Ollama and give it a
41. 1:48model to run. To do this, you go to the
42. 1:51Ollama site and you click on get
43. 1:53started. From there, you copy the
44. 1:55install command for whichever operating
45. 1:58system that you use. I use a MacBook, so
46. 2:00I'm going to go with Mac OS. I'll copy
47. 2:03the install command and then open a
48. 2:06where is it? Terminal window and paste
49. 2:09that command in, hit enter and as you
50. 2:11can see it will start downloading Ollama
51. 2:14for your desktop. That's it. Once it is
52. 2:17done, you can either run it by typing in
53. 2:20Ollama or as you can see, we now even
54. 2:23have it as a normal application that we
55. 2:25can just run and open from here. Then
56. 2:28you download a local model. Now, which
57. 2:31local model you download depends on
58. 2:33which computer you run. The more memory
59. 2:36or RAM your computer has, the bigger and
60. 2:39smarter model you can run. To know which
61. 2:42model you should download, check how
62. 2:44much RAM your computer has. So if I go
63. 2:47to my MacBook, I can see it has about 18
64. 2:51gigs of RAM. If you're on a normal
65. 2:53laptop with around 8 to 16 gigs of RAM,
66. 2:56then download Gemma 4 12b by just
67. 3:00selecting it. If you've got 18 to 32
68. 3:03gigs, then download Qwen 3 coder, pick
69. 3:06it and that's it. And if you've got 64
70. 3:09gigs or more, then you can download GPT
71. 3:13OSS 120, which is where the model will
72. 3:16feel a lot more like the paid thing. Oh,
73. 3:19and if you're on a Mac like I am, it's
74. 3:22important you know that your memory is
75. 3:24shared across everything. So if you're
76. 3:26on a Mac, pick a model that's about half
77. 3:28your memory size. If you have 18 gigs of
78. 3:32RAM like me, go for the models that work
79. 3:34for half of that size. So 9 gigs models
80. 3:37like Gemma 4 12b. And from there for
81. 3:40that model to actually download, you can
82. 3:42just send it hi and the first thing it
83. 3:44will actually do is start downloading
84. 3:46and installing it. For me, I've already
85. 3:48installed it, so I have access to it
86. 3:51right away. Now, I know figuring out
87. 3:53which model works best for your machine
88. 3:56sounds like a headache. So, I made a
89. 3:58free list that tells you exactly which
90. 4:01model to run. I've included it in the
91. 4:03first link inside the description. Just
92. 4:05find your RAM and it'll tell you what to
93. 4:08download. If you're an AI founder and
94. 4:10you're building something with AI, then
95. 4:12make sure to subscribe to that same
96. 4:14newsletter so you get every single drop
97. 4:17completely free of charge. But, when you
98. 4:20download whichever model is best for
99. 4:22you, you will now own that model for
100. 4:25free on your computer. And now that
101. 4:28we've downloaded our model, let me show
102. 4:30you how you can point Claude Code at it.
103. 4:32Pointing your local model at Claude Code
104. 4:34is now easier than ever. All you have to
105. 4:37do is, if you have your Ollama app still
106. 4:39open, click on apps and you'll see there
107. 4:42are two ways to run Claude Code here.
108. 4:45The first is by connecting it to the
109. 4:48Claude Code desktop app so that it
110. 4:50literally opens the Claude Code app for
111. 4:53you with the model of your choice. And
112. 4:55the second way is to click on Claude
113. 4:57Code in other apps. This will copy a
114. 5:00command into your clipboard so that all
115. 5:03you have to do is, again, go to your
116. 5:05terminal, paste that command in, click
117. 5:08launch or hit enter rather. And the
118. 5:10command is literally Ollama launch
119. 5:11Claude. Pick the model that we've
120. 5:13installed, which for me is Gemma 4 12B,
121. 5:16and it will start running and boom, just
122. 5:18like that. Look at that, Claude Code
123. 5:21Gemma 4 12B with high effort. You now
124. 5:24have the real Claude Code running on
125. 5:27your local model. Congratulations. And
126. 5:30that's the whole connection. Claude Code
127. 5:32now works exactly like it always has. It
128. 5:34just costs you nothing now. Now, let's
129. 5:37give it a job and test it. I'm going to
130. 5:39ask it to create a calculator app for us
131. 5:44and
132. 5:45open it. Now, I am filming this on my
133. 5:47MacBook Pro, which runs 18 gigs of
134. 5:50unified memory, which means I'm using a
135. 5:53Gemma 4 12B for this tutorial. And as
136. 5:56you will see, given this is a very small
137. 5:58local model, it will take its sweet time
138. 6:01until it finishes this job. But that's
139. 6:03what you need to understand about
140. 6:05something being free. Your machine will
141. 6:08have to do the heavy lifting instead.
142. 6:10And if you don't have a machine that
143. 6:12can, then the trade-off is your time.
144. 6:14Prompts take a lot longer to process.
145. 6:17But would you look at that? It finally
146. 6:20finished, so let's test it out. 2 + 2 is
147. 6:224, correct. * 8 is 32 / 8 is 4. There we
148. 6:27go. We have a calculator app. Running a
149. 6:30model locally is the private way to do
150. 6:33this, but it's not the only way to get
151. 6:36Claude Code for free. So, I thought I
152. 6:38would show you two more ways real quick.
153. 6:41The first one is a tool called Free LLM
154. 6:44API. Instead of running a model on your
155. 6:47machine, it stacks the free tiers almost
156. 6:50every AI lab gives you. Google, Grok,
157. 6:53Mistral, and a bunch more behind one
158. 6:57single endpoint, and it points Claude
159. 6:59Code at it. You bring your own free
160. 7:02keys, they stay encrypted on your
161. 7:04machine, and you get way more model
162. 7:06choice than you could ever run locally.
163. 7:09The trade-off is that it's not private
164. 7:11the way running local models is. Your
165. 7:14code still goes out to the providers,
166. 7:16but if you care about free more than you
167. 7:19care about private, this is probably the
168. 7:21best option for you. This is the repo
169. 7:23right here, and to install it, you first
170. 7:26have to install Docker. Pick whichever
171. 7:28version works on your operating system.
172. 7:31I'm on a Mac, so I would pick for Mac.
173. 7:33Then, once you set this up and it's
174. 7:35running, you open the Docker app, and
175. 7:38you go ahead and open a terminal window
176. 7:41inside of the Docker app, and then you
177. 7:43paste in this command. As a reminder,
178. 7:45all the commands, every single thing I'm
179. 7:48using is in the first link in the side
180. 7:50description below. And once it is done
181. 7:52installing, we grab the dashboard link,
182. 7:55which is over here, and we're going to
183. 7:57open it. Now, I'm already signed in, but
184. 8:00I want to show you what to do if you're
185. 8:02not. Since you're just going to be
186. 8:04installing this, it's going to ask you
187. 8:06for a setup code. What you have to do is
188. 8:09go back to the Docker and run Docker PS.
189. 8:12Then after Docker PS, you're going to
190. 8:15run this command, which if you notice
191. 8:17this section right here, we're inputting
192. 8:20with the name right here. So, those two
193. 8:23should be the same. If you press enter,
194. 8:26it doesn't show me because I've already
195. 8:27set it up, but it will give you the
196. 8:29setup code, which you will use to input
197. 8:32here, and boom, you're in. Then we go to
198. 8:35keys, we go to add key, and we pick
199. 8:38whatever free providers you want to add
200. 8:41from in the API keys. I recommend you go
201. 8:44with Google AI Studio. There's Grok as
202. 8:48well. There's Mistral. There are plenty,
203. 8:50but if we go with Google AI Studio, we
204. 8:52can click on get API key. If you don't
205. 8:55have an account, you can create one. You
206. 8:57go to create API key, you name it
207. 8:59whatever you want, say test one, create
208. 9:02a key, and it will give us an API key.
209. 9:04We will copy, we will paste it here, and
210. 9:07we will label this whatever we want. We
211. 9:09will add the key, press save, and boom,
212. 9:12just like that, we have that key. Then
213. 9:15inside your terminal app, you paste in
214. 9:17MPX free LMM API setup cloud. It will
215. 9:21ask you for the unified API key, which
216. 9:23you get from keys. You go to unified API
217. 9:26key, and over here, whoops, I almost
218. 9:29showed my key. You copy, and you paste
219. 9:32that in. Let me actually expand this
220. 9:34window, and I'm happy that we got this
221. 9:36error because if you get the same error,
222. 9:39it's pointing at a different localhost
223. 9:41than the one that we have. Ours, for
224. 9:44example, is 3001. It's pointing at
225. 9:46localhost 3000. So, instead, you can
226. 9:48paste this command instead, which points
227. 9:51it at the correct dashboard. Whichever
228. 9:54dashboard you have, if yours ends with
229. 9:553001, then put 3001. If it's different,
230. 9:58then put that. Enter. It's going to ask
231. 10:01us again for that same key. We're going
232. 10:03to go, copy it, go back in here, paste
233. 10:06it, hit enter, and boom. Just like that.
234. 10:09You're done. There, you can run Claude
235. 10:12in our terminal, and it will run using
236. 10:15Perplexity's free tiers instead. The
237. 10:17second option, and this is the one that
238. 10:20I recommend a lot more because it is a
239. 10:22lot easier than the one we just did, is
240. 10:25for you to install Open Code. It's a
241. 10:27free, open-source coding agent built to
242. 10:31run on any model you want from day one.
243. 10:33No setup tricks needed. Luckily,
244. 10:36installing Open Code is the easiest one
245. 10:38here. You just open Open Code, go to
246. 10:41download, pick whichever you install,
247. 10:44and you're here. With Open Code, you're
248. 10:45not running on Claude Code. However, it
249. 10:48is basically the same exact thing, minus
250. 10:50all the setup headache [clears throat]
251. 10:51that you just saw. And they give you
252. 10:53access to a bunch of free models from
253. 10:57the get-go without having to go through
254. 10:59any of that headache. Here, for example,
255. 11:01in this test, I asked it the same thing,
256. 11:04and it made us this app. Let me try to
257. 11:07find it. And there it is. It's literally
258. 11:09the same exact thing, if anything, a
259. 11:12little cleaner, in my opinion. And this
260. 11:15was one-shotted. So, that's the complete
261. 11:17free Claude Code map. If you want full
262. 11:20control and access to Claude Code for
263. 11:22free, you run it through Ollama. If you
264. 11:24want free Claude Code with more model
265. 11:27choices, then you run it through free
266. 11:29LLM API. And if you'd rather just switch
267. 11:33tools, then go with Open Code and skip
268. 11:35all the installation steps. Pick
269. 11:38whichever one fits you best. So, tell me
270. 11:40in the comments which one are you going
271. 11:43to try out first? Oh, and as I mentioned
272. 11:45in this video, the exact setup, every
273. 11:48command I ran and used, and the models
274. 11:51I'd actually pick are all listed for
275. 11:53free in the first link in the
276. 11:55description. If you enjoyed this video,
277. 11:57then leave a like. Or if you're new to
278. 11:59my channel, then subscribe because I've
279. 12:01got a ton more content just like this
280. 12:03coming your way. Oh, and would you look
281. 12:06at that? The algorithm gods have decided
282. 12:08you're very likely to enjoy this video
283. 12:11next, so click it, and I'll see you
284. 12:13there.