YouTube2Text

Stop Paying for Claude Code (3 Ways to actually use it for FREE!!) — Transcript

by Outskill · 5,668 words · 833 segments · language en · Watch on YouTube

Full transcript

  1. 0:00Claude Code is $20 a month, but not
  2. 0:03really. $20 [music] is what the AI
  3. 0:05inside it costs. Claude Code itself, the
  4. 0:08app, the tool, the thing that actually
  5. 0:10writes your code, is free. You just need
  6. 0:13to put a free engine inside it. Today,
  7. 0:15I'll show you three ways to do exactly
  8. 0:18that. [music] Let's build. Quick picture
  9. 0:20before we start. Think of a sports car.
  10. 0:22There's the body, the steering wheel,
  11. 0:24the dashboard, everything you interact
  12. 0:26with. And then, there's the [music]
  13. 0:28engine. Claude Code is the body. The AI
  14. 0:31is the engine. Today, we swap the engine
  15. 0:33three times. One runs [music] free on
  16. 0:35your laptop, one runs free in the cloud,
  17. 0:38and the third automatically switches
  18. 0:40between free AI models when you hit a
  19. 0:42limit.
  20. 0:47Okay, let's get into it. The first
  21. 0:50engine we're plugging in is called
  22. 0:51Ollama, and it's completely free. Just
  23. 0:55head over to ollama.com and hit
  24. 0:57download. It works on Mac, Linux, and
  25. 0:59Windows. I'm so I'll select Windows.
  26. 1:03Now, you can install it using a terminal
  27. 1:05command, or just download the installer.
  28. 1:07I'm going with the installer. Once it's
  29. 1:09downloaded, double-click the setup file
  30. 1:11and follow the installation steps. And
  31. 1:14that's it. Ollama is ready to use. Click
  32. 1:17the icon in the top left. Three things
  33. 1:19matter here. First, apps. Coding tools
  34. 1:22like Claude Code, Codex, Hermes, and
  35. 1:24others. Now, clicking one doesn't
  36. 1:26install anything. It just copies the
  37. 1:29command you need to connect Ollama to
  38. 1:30that tool. We'll use Claude Code in a
  39. 1:32minute. Second, chat. This is where we
  40. 1:35test models. Third, settings. The cloud
  41. 1:39toggle shows models on Ollama's servers.
  42. 1:42Those run on Ollama's servers. We're
  43. 1:44focusing on models that run directly on
  44. 1:46our laptop, so we can leave this off.
  45. 1:48Now, the one that actually matters,
  46. 1:51context length. Context is how much of
  47. 1:53your project the model can keep in its
  48. 1:55head at once. Claude Code works by
  49. 1:57reading your files, understanding them,
  50. 2:00then editing them. All of that has to
  51. 2:02fit inside the model's context. If it's
  52. 2:05too low, the model can lose track of
  53. 2:07your project as it works. Set it to 64K.
  54. 2:10That's the fix. Open chat and pick Qwen
  55. 2:123.54B,
  56. 2:14the small one, just to warm up the
  57. 2:15system. Type hi. First message always
  58. 2:18downloads the model, so it takes a
  59. 2:20second. After that, the model is already
  60. 2:23downloaded, so you won't have to
  61. 2:24download it again. But, hi doesn't prove
  62. 2:27anything. Time for the real test. Now,
  63. 2:29watch this. Wi-Fi off. Browser's dead.
  64. 2:33No internet on this machine at all. And
  65. 2:35I ask it to make a simple calculator
  66. 2:37app. And it writes it offline. And you
  67. 2:40don't have to worry about hitting a
  68. 2:42usage limit or waiting 5 hours,
  69. 2:44sometimes even a week, for it to reset.
  70. 2:46You can keep using it as much as your
  71. 2:48hardware allows. Every word of that code
  72. 2:51came from a model sitting on this
  73. 2:53laptop. And here's the biggest
  74. 2:55advantage, privacy. Your code and
  75. 2:58prompts stay on your computer. I'm
  76. 3:00switching the Wi-Fi back on for one
  77. 3:02reason, to paste the code into a free
  78. 3:04online compiler. And it runs. Is it
  79. 3:07pretty? No. But, a laptop with no
  80. 3:09internet just wrote working software. No
  81. 3:12subscription, no usage limits, nobody
  82. 3:15can switch it off. That's Engine 1. But,
  83. 3:17which model should you actually run?
  84. 3:20Don't memorize charts. Ask an AI. First,
  85. 3:23check your hardware. Press control plus
  86. 3:25shift plus escape, then click the
  87. 3:27performance tab. Mine says 12 GB of
  88. 3:29VRAM, that's your graphics card's
  89. 3:31memory, and 32 GB of RAM. Now, paste
  90. 3:34this into ChatGPT or any AI you have
  91. 3:37open. I have 12 GB of VRAM and 32 GB of
  92. 3:41RAM. I want to run a local model on
  93. 3:43Ollama for coding. Select the best model
  94. 3:45for my use case and reply with its exact
  95. 3:48Ollama name. It said Qwen 2.5 Coder 14B.
  96. 3:52Decent answer, but always ask a
  97. 3:54follow-up because the first answer might
  98. 3:56be outdated and new models keep coming
  99. 3:59out. Is this still the absolute best
  100. 4:01choice as of the current model releases.
  101. 4:03Check the latest available models before
  102. 4:05answering. And it corrected itself. Qwen
  103. 4:083.59B.
  104. 4:10That one extra question could save you
  105. 4:12from running a model that's already been
  106. 4:14replaced by something better. Back to
  107. 4:16Ollama to grab Qwen 3.59B.
  108. 4:18And here's the first problem. It's not
  109. 4:21in the drop down. Ollama only shows a
  110. 4:23selection of popular models in that
  111. 4:25menu. The model exists. It's sitting in
  112. 4:27Ollama's library, but to install it
  113. 4:30you'd have to find the exact command,
  114. 4:32open a terminal, and run it yourself.
  115. 4:34And that's where a simple setup can turn
  116. 4:36into an hour of Googling, Stack Overflow
  117. 4:39tabs, and random GitHub issues from
  118. 4:412023. But what if you didn't have to
  119. 4:43figure out any of that yourself? That's
  120. 4:46where another AI comes in. The one I'm
  121. 4:48using is called Anti-Gravity, a free
  122. 4:51coding tool from Google. What makes it
  123. 4:53different from a normal chatbot is that
  124. 4:55it can actually work on your computer.
  125. 4:57It opens your terminal, runs commands,
  126. 5:00reads error messages, and even fixes
  127. 5:02problems along the way. So you tell it
  128. 5:04in plain English what you want
  129. 5:05installed, and it just does it. To use
  130. 5:07it, download Anti-Gravity, sign in with
  131. 5:09a Google account, open it up, and give
  132. 5:11it one sentence. I am running AI models
  133. 5:14locally. Install the model Qwen 3.59B
  134. 5:18and set up everything so it works with
  135. 5:20Ollama. It finds the model, downloads
  136. 5:22it, hooks it into Ollama, done. One
  137. 5:25thing to watch. Anti-Gravity will ask
  138. 5:27for permission a few times while it
  139. 5:29works because it's running commands on
  140. 5:31your machine. Read each prompt before
  141. 5:33you click allow. You're the driver. It's
  142. 5:35the mechanic. When it's finished, close
  143. 5:38Ollama and reopen it. The new model
  144. 5:40shows up in the drop down. And this is
  145. 5:42the house rule for the whole video.
  146. 5:44Anything longer than one line, you don't
  147. 5:46type it yourself. An AI installs it. If
  148. 5:49something breaks on the way, copy the
  149. 5:51error, paste it back into the chat, and
  150. 5:53say, "Fix this." Now for the part you
  151. 5:55came here for. Back to Ollama, click the
  152. 5:58apps tab, and click Claude code. See
  153. 6:00that little flash in the corner? That's
  154. 6:02the install command for Claude code
  155. 6:04copied to your clipboard. Make a folder
  156. 6:06anywhere on your computer. Mine is
  157. 6:08called free Claude code. Open it, right
  158. 6:11click in the empty space, choose open in
  159. 6:13terminal, paste, and hit enter. If it'll
  160. 6:16say Claude code is not installed,
  161. 6:18install now? Type Y. And this is
  162. 6:20important. What's installing right now
  163. 6:23is the real terminal-based Claude code,
  164. 6:25Anthropic's genuine app, straight from
  165. 6:27its official source. Not a copy, not a
  166. 6:30clone, not a wrapper. Ollama just runs
  167. 6:32the official installer for you. When it
  168. 6:34finishes, it asks which model to use.
  169. 6:37Pick Claude 3.59B.
  170. 6:39And there it is, Claude code sitting in
  171. 6:41your terminal with a free engine inside
  172. 6:44it. No Anthropic subscription, no API
  173. 6:46key. You're running it with a local
  174. 6:48model through Ollama. Let's put it to
  175. 6:50work. Type, "Design me a simple
  176. 6:53website." Or, if you want to keep it
  177. 6:55even simpler, "Build me a calculator
  178. 6:57app." It'll ask you a question or two to
  179. 6:59understand what you want, then ask
  180. 7:01permission to create files in the
  181. 7:02folder. Allow it. First draft is going
  182. 7:05to be rough, honestly. But, that's the
  183. 7:07whole point. You iterate. Make the
  184. 7:09buttons bigger, switch to dark mode,
  185. 7:11clean up the layout. Each round, it gets
  186. 7:13closer. And remember, this is all
  187. 7:15running locally, with no cloud AI usage,
  188. 7:18no provider imposed usage limits, and it
  189. 7:21still works with the Wi-Fi off. Next
  190. 7:23time you want to use it, you don't
  191. 7:24repeat any of this setup. Right click
  192. 7:26your folder, open in terminal, and type
  193. 7:29one line. That's the whole routine from
  194. 7:31now on. So, why isn't this video over?
  195. 7:34Because you saw it, too. It's slow. And
  196. 7:36honestly, it's not that smart. So, let's
  197. 7:39upgrade. The next engine runs in the
  198. 7:41cloud. Faster, smarter, still free. And
  199. 7:44this is where the model menu starts
  200. 7:46lying to you. The second engine is
  201. 7:48called OpenRouter. Think of it as a
  202. 7:50middleman that gives you access to
  203. 7:52dozens of AI models through one single
  204. 7:54key. And a bunch of those
  205. 7:56[clears throat] models are completely
  206. 7:57free. Nothing runs on your laptop this
  207. 7:59time. The AI lives on their servers. You
  208. 8:02just need one key to talk to it. Head to
  209. 8:05openrouter.ai and sign up. Free, no
  210. 8:08credit card. Once you're in, click API
  211. 8:10keys, then create key. Give it a name.
  212. 8:13Mine is Claude code, and set an
  213. 8:15expiration date. I set mine to 7 days,
  214. 8:18which means by the time you're watching
  215. 8:20this, my key is already dead. If you
  216. 8:22ever add money to this account later,
  217. 8:24set a spending limit, too. It's a
  218. 8:26seatbelt, wear it. Create the key and
  219. 8:29copy it somewhere safe right now,
  220. 8:31because Openrouter only shows it to you
  221. 8:33once. And treat it like a password.
  222. 8:36Anyone who gets this key can burn
  223. 8:37through your quota. Before we plug this
  224. 8:39key into anything, look at what you just
  225. 8:42unlocked. Click models at the top, open
  226. 8:44the price filter, and set it to free.
  227. 8:47Look for models with the colon free
  228. 8:49label. Those are the ones you can use
  229. 8:50without buying credits. Every request to
  230. 8:53those models costs zero. Now, here's the
  231. 8:55catch. This list rotates. Models join,
  232. 8:59models leave, companies pull their free
  233. 9:01tiers all the time. So, the names you
  234. 9:03see today might not be the names you see
  235. 9:05next month. The names don't matter. The
  236. 9:08filter does. As long as you know how to
  237. 9:10find the free list, you'll always have
  238. 9:12an engine. Now, we plug this key into
  239. 9:14Claude code. The block of code you need
  240. 9:16is sitting in the description of this
  241. 9:18video. Copy it. You only change one
  242. 9:20thing in it. Your key goes between these
  243. 9:22two quotes. The rest of the block is
  244. 9:25just instructions that tell this
  245. 9:26specific terminal window to send all its
  246. 9:29work to Openrouter instead of Anthropic.
  247. 9:31One of those lines clears out the old
  248. 9:33key slot, so there's zero chance of
  249. 9:35accidentally falling back to a paid key.
  250. 9:38And these three lines near the bottom,
  251. 9:39the ones labeled Opus, Sonnet, and
  252. 9:41Haiku, I added those myself on purpose.
  253. 9:45Here's why. Claude code's model menu has
  254. 9:47three seats inside it, named after
  255. 9:49Anthropic's paid models. These lines map
  256. 9:52each of those seats to a free model on
  257. 9:54Openrouter. So, whichever seat you pick
  258. 9:56inside Claude code, the model we assign
  259. 9:59to that seat is the one answering. Hold
  260. 10:01on to that. It matters in a minute. Copy
  261. 10:03the whole block. Back to the folder we
  262. 10:05made earlier, right click, open in
  263. 10:08terminal, paste, hit enter. Claude code
  264. 10:11opens up again. Same app you were using
  265. 10:135 minutes ago, completely different
  266. 10:15engine behind it. Say hi. The reply
  267. 10:17comes back before you finish blinking.
  268. 10:19That's the difference between a model
  269. 10:21running on your laptop and a model
  270. 10:23running in the cloud. And if you want
  271. 10:25proof of where the work is actually
  272. 10:26going, type {slash} status. See that
  273. 10:29address? That's OpenRouter. That's where
  274. 10:31every message you send from now on is
  275. 10:33traveling. Now, remember the model menu
  276. 10:36I teased earlier? Type {slash} model.
  277. 10:38Look at that. Sonnet, Haiku. Pick
  278. 10:41Sonnet. Ask it literally anything. It
  279. 10:44answers. So, did we just get Claude
  280. 10:47Sonnet for free? Honestly, no. And I'll
  281. 10:50prove it in one screenshot. Open
  282. 10:52openrouter.ai/activity
  283. 10:55in your browser. This is the real log.
  284. 10:57Every request you've sent, the exact
  285. 11:00model that actually answered it, and the
  286. 11:02price. All zeros. And nowhere on this
  287. 11:05page does the word Claude appear. The
  288. 11:07menu inside Claude code shows seat
  289. 11:09names. OpenRouter's activity log shows
  290. 11:11the actual passenger. Any video out
  291. 11:14there calling this free Sonnet is
  292. 11:16reading the seat label and not checking
  293. 11:18who's actually in the seat. One more
  294. 11:20thing that might scare you if you don't
  295. 11:22know about it. Open Claude code's own
  296. 11:24usage screen and it'll tell you this
  297. 11:26session cost you around 45 cents. Relax.
  298. 11:30That number may be an estimate based on
  299. 11:32Anthropic's pricing, not what OpenRouter
  300. 11:34actually charged you. So, don't judge
  301. 11:36the bill by that screen. The number
  302. 11:38that's actually real is the one on
  303. 11:40OpenRouter. And on OpenRouter, you paid
  304. 11:43zero. But, there's a trade-off. Unlike
  305. 11:46the local setup, your prompts and
  306. 11:48project files are now being sent to
  307. 11:50external servers. So, you need to be
  308. 11:52careful about what you share. And you're
  309. 11:54also dependent on their servers, their
  310. 11:56free models, and whatever usage limits
  311. 11:59they impose. Okay, so the same task that
  312. 12:01took the local engine a few minutes, the
  313. 12:03cloud writes it in seconds. But, here's
  314. 12:06the honest answer. OpenRouter's free
  315. 12:09tier is capped. For models marked
  316. 12:11{colon} free, you're limited to 20
  317. 12:13requests a minute and 50 requests a day
  318. 12:16on a new account.
  319. 12:17>> 50 sounds like plenty until you remember
  320. 12:19what Claude code actually does. One task
  321. 12:22from you can trigger 10, 15, 20 requests
  322. 12:25behind the scenes because it's reading
  323. 12:27files, writing files, checking its own
  324. 12:29work, then fixing what it broke. One
  325. 12:32real coding session can empty your whole
  326. 12:34daily cap. And [snorts] when you hit
  327. 12:36zero, you get error 429, rate limited,
  328. 12:39mid task, frozen, come back tomorrow.
  329. 12:42That wall right there is the exact
  330. 12:44reason the third method exists. The
  331. 12:46third setup is called Omni route. Here's
  332. 12:49what it actually does. It's a free app
  333. 12:51that connects Claude code to over 375
  334. 12:54providers and gives you access to more
  335. 12:57than 1,400
  336. 12:58AI models, including Claude, GPT,
  337. 13:01Gemini, Kimi, and Deep Seek. Open router
  338. 13:04has a free cap. Fine, but there are
  339. 13:06other services out there with their own
  340. 13:08free caps, Nvidia, Hero, Kline, a bunch
  341. 13:12of others. On their own, each one is
  342. 13:14small. Together, they're huge. And
  343. 13:17that's where Omni route comes in.
  344. 13:19>> [clears throat]
  345. 13:19>> It brings all of them together in one
  346. 13:21place. When one provider runs out of
  347. 13:24free usage, it can switch to another one
  348. 13:26automatically. You don't have to keep
  349. 13:28changing providers yourself. One AI runs
  350. 13:30out of fuel, another one takes over. And
  351. 13:33that's what we're setting up today.
  352. 13:34We're going to install it the same way
  353. 13:36we fixed Ollama earlier. Anti-gravity
  354. 13:38does the work. Open Anti-gravity, paste
  355. 13:41this prompt from the description, and
  356. 13:43let it go. It'll install Claude code if
  357. 13:45you don't have it already, install Omni
  358. 13:47route, start the service, and open the
  359. 13:49dashboard in your browser. While it's
  360. 13:51working, it'll ask for permission more
  361. 13:53than three times. Read each one before
  362. 13:56you click allow. It's running real
  363. 13:58commands on your machine. Once the
  364. 13:59installation finishes, click the
  365. 14:01localhost link in the dashboard. Then
  366. 14:04the little arrow at the top right opens
  367. 14:06it in your browser. First time you log
  368. 14:08in, the default password is just the
  369. 14:10word change me. Paste it in and change
  370. 14:12it to something real in settings right
  371. 14:14after. All right, now the fun part.
  372. 14:17We're going to start connecting engines
  373. 14:18to this thing. On the dashboard, open
  374. 14:21the Omni proxy tab. This is where you
  375. 14:23plug in every free service you want Omni
  376. 14:25route to pull from. We'll do five of
  377. 14:27them. First up, Kiro. Click the
  378. 14:29providers option, search Kiro, accept
  379. 14:32the warning that pops up, and sign in
  380. 14:34with your free Kiro account. Done. Now
  381. 14:36look at its model list. Yes, those are
  382. 14:39real Claude models. Sonnet, Haiku, the
  383. 14:42actual ones. But here's something you
  384. 14:44need to know before you get too excited.
  385. 14:46Kiro's free tier is 50 credits a month,
  386. 14:49which is maybe a handful of prompts
  387. 14:50before it's gone. And Omni route itself
  388. 14:53puts a warning on the screen saying
  389. 14:55Kiro's terms of service don't
  390. 14:56technically allow this kind of use. So
  391. 14:59use it at your own risk. I'm showing you
  392. 15:01how it works, but whether you want to
  393. 15:03connect your account is entirely your
  394. 15:05call. Second, the two that take a key.
  395. 15:08Nvidia [clears throat]
  396. 15:09is the backbone of this setup. Create a
  397. 15:11free API key at build. nvidia.com. Paste
  398. 15:14it into the Nvidia card, click check,
  399. 15:16and set its priority to one so it's used
  400. 15:19first. Nvidia mixes free and paid
  401. 15:21models, so don't worry about
  402. 15:23accidentally using a paid one just yet.
  403. 15:25We're going to filter those out in a
  404. 15:26minute. Then open router. Paste the same
  405. 15:29free key from method two. Same free
  406. 15:31models, same limit, 50 requests a day.
  407. 15:34Next, the two that just need a sign in.
  408. 15:37Klein, an AI coding assistant whose new
  409. 15:39models are often free for a while. Sign
  410. 15:41in with a free Klein account. And
  411. 15:43anti-gravity. Sign in with Google to
  412. 15:45unlock Gemini and some Claude models
  413. 15:47within a usage limit. Same warning as
  414. 15:50Kiro though. Google's terms don't allow
  415. 15:52this, and Omni route flags it right
  416. 15:54here. Your account, your call. Back on
  417. 15:56the home dashboard, mine shows four
  418. 15:58connected. Klein threw an error, but
  419. 16:00four is plenty. And stay with me because
  420. 16:02in about two minutes, we're going to get
  421. 16:04the real Claude to answer through this
  422. 16:06exact setup. Now we put all of this into
  423. 16:09one combo. A combo is just Omni route's
  424. 16:12word for a pool of engines that work
  425. 16:14together. Click the combos tab, hit
  426. 16:16create your first combo, and give it a
  427. 16:18name. Mine is Claude. Under steps, you
  428. 16:21pick the first provider, the model to
  429. 16:23use, and the account. I'm starting with
  430. 16:26GPT-OSS-120B
  431. 16:28on Nvidia. Click add. Then Omni Route
  432. 16:30asks you how it should pick between all
  433. 16:32your engines when a request comes in.
  434. 16:34There are multiple strategies. I'm going
  435. 16:37with intelligent auto. It makes the
  436. 16:39decision for me every single time, which
  437. 16:41is exactly what we want. The whole point
  438. 16:43is to stop manually switching between
  439. 16:45providers. Alternatively, you could also
  440. 16:48choose round robin or any other you
  441. 16:50like. Next, in the candidate pool below,
  442. 16:52pick every provider you connected. Back
  443. 16:55in steps, add a second engine, Open
  444. 16:57Router this time as the backup. The up
  445. 17:00and down arrows next to each engine set
  446. 17:02the order they get tried in. Top one
  447. 17:04goes first. Hit create. Hit save. And
  448. 17:07just like that, you've got your engine
  449. 17:09pool ready to go. One warning before we
  450. 17:11move on. When you add Open Router as
  451. 17:14that second engine, it'll ask you which
  452. 17:16specific model to use from Open Router.
  453. 17:18And one of the options in that drop down
  454. 17:20is also called auto. Normally, you don't
  455. 17:23want to pick that one. Open Router's
  456. 17:26auto can quietly route your request to a
  457. 17:28paid model and charge you. But for now,
  458. 17:30just pick it. We'll fix that in a
  459. 17:32moment, so you don't have to worry about
  460. 17:34accidentally getting charged. Now we
  461. 17:36need to connect Claude Code to this
  462. 17:38engine pool we just built. For that,
  463. 17:40Omni Route gives you its own API key.
  464. 17:42Think of it as the master key that
  465. 17:44unlocks every engine inside the combo.
  466. 17:47Head to the API keys tab inside Omni
  467. 17:49Route, hit create, copy the key, and
  468. 17:51save it somewhere safe. Now flip back to
  469. 17:54Anti-Gravity one more time. There's a
  470. 17:56prompt in the description called
  471. 17:57shortcuts. Paste it in. This prompt does
  472. 18:00three things at once. It takes your Omni
  473. 18:02Route key and saves it into a config
  474. 18:04file. It creates a dedicated work folder
  475. 18:07on your computer, and it drops two
  476. 18:09shortcuts right on your desktop. One
  477. 18:11that opens Claude Code already wired up
  478. 18:13to Omni Route, and one that opens the
  479. 18:15Omni Route dashboard in your browser.
  480. 18:17Halfway through the install,
  481. 18:19Anti-Gravity will stop and ask you for
  482. 18:21the key. Important, don't paste it into
  483. 18:23the chat. That's the fastest way for a
  484. 18:25key to leak. Instead, it'll open a file
  485. 18:28for you. Paste the key into that file,
  486. 18:30save it, and type continue in the chat.
  487. 18:33If the file doesn't open on its own,
  488. 18:35just say open notepad so I can paste the
  489. 18:37key. Done. Now check your desktop. Free
  490. 18:40Claude code and Omni route dashboard.
  491. 18:42Two icons ready to go. But before you
  492. 18:44double click anything, there's one more
  493. 18:46trap you need to know about. Each
  494. 18:48service we connected, Hero, Nvidia,
  495. 18:51OpenRouter, all of them, has both free
  496. 18:53and paid models on their platform. And
  497. 18:56when Claude code's model menu opens up,
  498. 18:58it shows you all of them mixed together
  499. 19:00with no label telling you which is
  500. 19:02which. So you could accidentally pick a
  501. 19:04paid model, send one message, and end up
  502. 19:06with a bill or an error message if you
  503. 19:08have no credits. To fix it, there's one
  504. 19:11more prompt in the description. I called
  505. 19:13it the free models prompt. Paste it into
  506. 19:15anti-gravity. It goes through every
  507. 19:17connected provider, filters out every
  508. 19:19paid model from the menu, leaves only
  509. 19:22the free ones visible, and runs a test
  510. 19:24to confirm the fix held. From that point
  511. 19:26on, the models you see in that menu are
  512. 19:28the ones we've configured as free. No
  513. 19:30more guessing. All right, moment of
  514. 19:33truth. Double click the free Claude code
  515. 19:35shortcut on your desktop. Terminal
  516. 19:37opens, Claude code loads, say hi. It
  517. 19:40replies. Now ask it what model it is,
  518. 19:43and it tells you Claude. But you know
  519. 19:45the rule from method two, a model's word
  520. 19:48isn't proof. Some free models will even
  521. 19:50claim to be Claude. So check the
  522. 19:52receipt. Open the OmniRouter dashboard,
  523. 19:55go to logs, and look at the last
  524. 19:56request. Provider, Hero. Model, Claude
  525. 20:00Sonnet. That's the real Claude Sonnet
  526. 20:02talking to you right now, routed through
  527. 20:04Hero's free credits. Hero's free credits
  528. 20:07are tiny, so this won't last long. When
  529. 20:09they run out, switch to the Pombo we
  530. 20:11built. Inside it, Nvidia and OpenRouter
  531. 20:14take over on their own. Just know that
  532. 20:16from then on, it isn't Claude answering.
  533. 20:18It's the free engines. You keep working,
  534. 20:20and the switch happens behind the
  535. 20:22scenes, and the bill stays at zero. All
  536. 20:25right, let's put this whole setup
  537. 20:26through a real test, not a toy prompt,
  538. 20:29something you'd actually want to build.
  539. 20:31I typed in, "Design me a simple personal
  540. 20:33note-taking application." And watch what
  541. 20:36happens here, because this is the part
  542. 20:38most tutorials skip. Claude Code doesn't
  543. 20:40just start generating random code. It
  544. 20:42stops and asks me three questions first.
  545. 20:45Do I want a desktop app or a web app? Do
  546. 20:47I want my notes stored locally on my
  547. 20:49machine or in the cloud? And which
  548. 20:51features matter most to me? Tags,
  549. 20:54search, markdown, dark mode? I answer
  550. 20:57each one, it builds a plan based on my
  551. 20:59answers, shows me the plan, and only
  552. 21:01then starts coding. A few minutes later,
  553. 21:04I've got a working note-taking app
  554. 21:05running on my own computer on localhost.
  555. 21:08Open it in a browser, create a note,
  556. 21:10save it, close the app, reopen it, note
  557. 21:13still there. Fully working, zero dollars
  558. 21:16spent. Now, I want to be straight with
  559. 21:18you about something. If you've searched
  560. 21:20for free Claude videos before, you've
  561. 21:22probably seen a hundred titles screaming
  562. 21:25"Unlimited free Claude." It's not true.
  563. 21:28There is no setup on the internet right
  564. 21:30now that gives you unlimited access to
  565. 21:32the real Claude model for free. What you
  566. 21:34actually get with this build is two
  567. 21:36things stacked together. A small taste
  568. 21:38of real Claude through Keiro's monthly
  569. 21:40credits, plus the Claude Code tool
  570. 21:43itself running completely free with
  571. 21:45other engines inside it. That's the
  572. 21:47honest version. And here's the thing,
  573. 21:49for most people that stack is already
  574. 21:52enough. You get real Claude for the hard
  575. 21:54problems where you need its brain, and
  576. 21:56you get free open-source engines for
  577. 21:58everything else. You don't have to stop
  578. 22:00working just because one provider runs
  579. 22:02out of credits. You never see a bill as
  580. 22:05long as you stick to the free models and
  581. 22:07usage limits we've configured. And
  582. 22:09that's the real advantage of this setup.
  583. 22:11Not unlimited Claude, but a free coding
  584. 22:14workflow that keeps going even when one
  585. 22:16engine runs out. One bonus before we
  586. 22:18wrap up. If you don't specifically need
  587. 22:21Claude Code itself and just want to code
  588. 22:23every day with the best free AI models
  589. 22:25possible, there's a simpler setup. And
  590. 22:28honestly, it's the one I'd keep coming
  591. 22:30back to. It's called Kilocode and it
  592. 22:32lives inside VS Code, Microsoft's free
  593. 22:35code editor. One honest note before we
  594. 22:37start. This isn't Claude Code. It's a
  595. 22:40different coding agent made by a
  596. 22:42different team. But the idea is exactly
  597. 22:45the same as everything we just built. A
  598. 22:47free engine inside a coding tool. Think
  599. 22:49of this as the lightweight daily driver.
  600. 22:52Claude Code when you want the real deal,
  601. 22:54Kilocode when you just want to ship
  602. 22:55stuff without thinking about
  603. 22:57subscriptions. All right, let's build
  604. 22:59this one from scratch. Head to
  605. 23:01code.visualstudio.com
  606. 23:03and download VS Code. Install it, open
  607. 23:05it up, and the first screen asks you to
  608. 23:07pick a theme. I went with dark, but pick
  609. 23:10whatever you want. Once you're in, click
  610. 23:12open folder at the top left, choose any
  611. 23:15folder on your computer where you want
  612. 23:16to work, and when it asks if you trust
  613. 23:18the files inside, say yes. Now we add
  614. 23:21Kilocode. On the left side bar, click
  615. 23:23the extensions icon. It looks like four
  616. 23:26little squares. Search Kilocode, click
  617. 23:28install, and give it a second. Once it's
  618. 23:31installed, a new icon appears in that
  619. 23:33same side bar. Click it. Kilo asks you
  620. 23:36to sign in. Do that and you're done
  621. 23:38setting up. That's it. You now have a
  622. 23:40coding agent running inside VS Code and
  623. 23:43you haven't paid for anything. Open
  624. 23:45Kilo's model menu for a second. Right
  625. 23:47out of the box, it already comes with
  626. 23:49free models preloaded, Auto Free, Nvidia
  627. 23:52Neumotron, and a few others. So
  628. 23:54technically, you could start coding
  629. 23:56right now without connecting anything
  630. 23:58else. But this is Kilo's own curated
  631. 24:01free tier and the lineup changes every
  632. 24:03few weeks as deals come and go. So
  633. 24:05instead of relying only on what Kilo
  634. 24:07gives you, we're going to add our own
  635. 24:09providers underneath. That way, you've
  636. 24:12always got multiple engines on tap. And
  637. 24:14when one hits its limit, you can switch
  638. 24:16to another. Let's start with Nvidia, the
  639. 24:18same provider we used in Omni route. Go
  640. 24:21to Kilo settings, click providers, then
  641. 24:23custom providers, and hit connect.
  642. 24:26There's a block of connection details in
  643. 24:27the description. Paste it in. Then, grab
  644. 24:30your Nvidia API key, the same one you
  645. 24:32created earlier or a new one if you
  646. 24:34prefer, and paste it into the key field.
  647. 24:37When you hit submit, Killo pulls down
  648. 24:39the full list of every model Nvidia
  649. 24:41offers. It's long, and you don't need
  650. 24:44all of them. We're just going to pick
  651. 24:46three. Click deselect all at the top,
  652. 24:48then individually tick these three:
  653. 24:50Nemotron 3 Ultra, GLM 5.3, and GLM 5.3
  654. 24:55Flash. These are the three free models
  655. 24:58we're using for this setup. Below the
  656. 25:00list, turn on reasoning so the models
  657. 25:02can think through longer problems, and
  658. 25:04turn on images only for models that can
  659. 25:07actually read images. If you flip it on
  660. 25:09for a text-only model, it'll error out.
  661. 25:11Hit submit. Nvidia is now live inside
  662. 25:14Killo, and honestly, Nvidia is my
  663. 25:17favorite provider in this whole video.
  664. 25:19Its free tier is generous, with a limit
  665. 25:21of around 40 requests a minute according
  666. 25:23to the current offering. That's plenty
  667. 25:25for testing and personal projects, which
  668. 25:27is exactly what we're doing. All right,
  669. 25:29let's test it with something real. I
  670. 25:32opened Killo, picked GLM 5.3 Flash from
  671. 25:35the model drop-down, the fastest of the
  672. 25:37three Nvidia models we just added, and
  673. 25:39gave it one prompt. It's in the
  674. 25:41description. Just a short list of what I
  675. 25:43wanted: a start screen, enemy waves,
  676. 25:45power-ups, a boss, and a high score. And
  677. 25:48watch what happens. It wrote the entire
  678. 25:51game. One file, around 1,200 lines of
  679. 25:54code. Double-click it, and the game runs
  680. 25:56in your browser. Start screen with
  681. 25:58instructions, enemy waves that get
  682. 26:00harder as you go, power-ups that drop
  683. 26:02from enemies, and a boss fight every
  684. 26:05fifth wave. And this isn't just a
  685. 26:07screenshot of a game. I can actually
  686. 26:09play it. I played it for about 10
  687. 26:11minutes while editing this video, and
  688. 26:12the whole thing, concept to playable,
  689. 26:15cost me zero dollars. That's the kind of
  690. 26:17thing you can start doing with these
  691. 26:19free models today. Now, that free Nvidia
  692. 26:22tier is generous, but it's still a tier.
  693. 26:25You will hit the limit eventually,
  694. 26:27especially on bigger projects. And this
  695. 26:29is where the setup gets useful. Instead
  696. 26:32of waiting for your limit to reset, you
  697. 26:34just add another engine. The setup is
  698. 26:36the same every time. Open settings,
  699. 26:38providers, custom providers, connect,
  700. 26:41and repeat the exact same flow we just
  701. 26:43did for Nvidia, but with different
  702. 26:45services. First, open router. Paste in
  703. 26:48the same key you created earlier, and
  704. 26:50this time you've got access to its free
  705. 26:52models like Nemotron 3 Ultra, Inkling,
  706. 26:55and Gemma 4. Limit, 50 requests a day.
  707. 26:58Next, Gemini. Head to Google AI Studio,
  708. 27:01generate a free key, and paste it in.
  709. 27:03The connection details are in the
  710. 27:05description. One important thing to flag
  711. 27:07about Gemini, on the free tier
  712. 27:09specifically, Google may use your
  713. 27:11prompts to improve its products. So, if
  714. 27:13you're writing anything sensitive,
  715. 27:15client code, proprietary logic, anything
  716. 27:17you don't want being used for model
  717. 27:19improvement, don't route it through
  718. 27:21Gemini's free tier. Use it for personal
  719. 27:23projects instead. And last one, Kline.
  720. 27:26This one isn't added inside Kilo. It's
  721. 27:28its own extension, and you don't need an
  722. 27:30API key. Install the Kline extension
  723. 27:32inside VS Code, sign in with their
  724. 27:34account, and you're set. The interesting
  725. 27:37thing about Kline is that it
  726. 27:38periodically changes which free models
  727. 27:40it offers. So, the lineup you see today
  728. 27:42might be different next week. And there
  729. 27:44are plenty more free providers out
  730. 27:46there. I've listed them all in the
  731. 27:48description. Now, here's a question
  732. 27:50you're probably going to run into
  733. 27:51eventually. What happens when you're
  734. 27:53mid-task, mid-code generation, and the
  735. 27:56model you're using suddenly hits its
  736. 27:58daily limit? Kilo handles this pretty
  737. 28:00keenly. Inside the same chat thread,
  738. 28:03just open the model drop-down at the
  739. 28:04bottom and pick a different model. Kilo
  740. 28:07carries forward the available
  741. 28:08conversation and project context, so the
  742. 28:11new model can pick up where you left
  743. 28:12off. You don't have to start explaining
  744. 28:14the entire project from scratch. And if
  745. 28:17you're starting a brand new chat and you
  746. 28:18want Kilo to remember your project long
  747. 28:20term, what you're building, what's done,
  748. 28:23and what's next, turn on its memory
  749. 28:25feature by typing {slash} memory on.
  750. 28:27From that point on, Kilo keeps running
  751. 28:29notes about your project across
  752. 28:31conversations. So, when you switch
  753. 28:33models or start a new chat, that saved
  754. 28:35project context can help you get back to
  755. 28:38work without starting from zero. So,
  756. 28:40we've built four completely different
  757. 28:42setups. And the real question is, which
  758. 28:44one should you actually use day-to-day?
  759. 28:47Here's the honest answer. If you've got
  760. 28:49a decent graphics card and privacy
  761. 28:51matters to you, especially if you're
  762. 28:53working with code you don't want sitting
  763. 28:55on someone else's server, go with
  764. 28:57Ollama. It works offline, but you're
  765. 28:59limited by your own hardware. If you've
  766. 29:01got no graphics card and just want speed
  767. 29:03today with no hardware investment,
  768. 29:06OpenRouter is your pick. Just remember
  769. 29:08the 50 requests a day cap on the free
  770. 29:10models. If you want the real Claude code
  771. 29:12experience with multiple free engines
  772. 29:14behind it, Omni route is the one to try.
  773. 29:17And if you don't specifically need the
  774. 29:19Claude code interface and just want a
  775. 29:21straightforward way to use free coding
  776. 29:23models, Kilo code inside VS Code is
  777. 29:26probably the easiest daily driver. And
  778. 29:28whichever one you choose, there are four
  779. 29:30things you need to know before you start
  780. 29:32using it seriously. Here they are.
  781. 29:34One, nothing is unlimited. Every free
  782. 29:37tier has a cap, and real Claude is never
  783. 29:40free for long. The whole point of this
  784. 29:42setup is to combine those limits, not
  785. 29:44pretend they don't exist. Two, free
  786. 29:47cloud engines see what you send. Client
  787. 29:50code, passwords, private files, anything
  788. 29:52sensitive, keep those on Ollama. Free
  789. 29:55doesn't automatically mean private.
  790. 29:58Always check what a provider does with
  791. 29:59your data before sending anything
  792. 30:01sensitive. Three, some connections may
  793. 30:04violate the provider's terms. Kiro and
  794. 30:06Anti-Gravity are flagged right on
  795. 30:08screen. Your account, your call. I'm
  796. 30:11showing you how these connections work,
  797. 30:13but you should always check the terms
  798. 30:14yourself before using them. Four, expect
  799. 30:17errors. We hit plenty. Copy the error,
  800. 30:21paste it into Anti-Gravity and say, "Fix
  801. 30:23this." That's part of working with AI
  802. 30:25coding tools. Sometimes the first
  803. 30:27attempt fails, but you can usually use
  804. 30:29the error message to figure out what
  805. 30:31went wrong. And one final rule, whatever
  806. 30:34a menu tells you, check the receipt, the
  807. 30:36actual usage and billing page, not the
  808. 30:39label. Everyone told you Cloud Code
  809. 30:41costs $20 a month. Today, we got the
  810. 30:44real Cloud Code, Anthropic's actual
  811. 30:47coding app running with free engines
  812. 30:49inside it three different ways. One
  813. 30:52works with the Wi-Fi off. One lets you
  814. 30:54use Cloud models without investing in
  815. 30:56expensive hardware. One pools multiple
  816. 30:58free engines together so you can switch
  817. 31:00when you hit a limit. And on top of all
  818. 31:02that, you've got a bonus daily driver
  819. 31:04inside VS Code that just built a working
  820. 31:07arcade game from a single prompt. No
  821. 31:09subscription, no credit card, no
  822. 31:11Anthropic login. Every prompt from this
  823. 31:14video, every base [music] block, every
  824. 31:16link to the tools we used, it's all
  825. 31:18inside my free Stay Ahead community.
  826. 31:20Link's in the description. Join, grab
  827. 31:22everything, and run your own build
  828. 31:24tonight. And tell us in the comments,
  829. 31:26which setup are you trying first?
  830. 31:28>> [music]
  831. 31:28>> Our team actually reads them, and we try
  832. 31:30to pick the next video based on what you
  833. 31:32ask for.

About this transcript

This page contains the full transcript of Stop Paying for Claude Code (3 Ways to actually use it for FREE!!) by Outskill, generated from the public captions YouTube serves with the video. The transcript has 5,668 words across 833 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.