Anthropic started localizing Claude pricing outside the US. India (its second-largest market, ~5.8% of global Claude usage) got rupee pricing today: Claude Pro starts at ₹1,999/mo (~$21), Max 20x up to ₹23,999/mo, tax included. No UPI yet (card or app-store billing only).
Read-through: this is a defensive move; OpenAI and Google already had local pricing/payment options. When labs fight on local currency + tax, that floor pressure eventually reaches your API sticker. Don't lock pricing to one vendor's number.
A system, not a freelancer: a 30-year operator (real estate, hospitality, food, a theater) directing a full team of AI agents at one client project at a time, strategy to ship. You own the code and the keys. When Claude's price changes by country, you want someone steering the whole fleet. Book a 30-min call. https://outpace.media
The stories
LEAD (builder impact)
— Claude Code shipped /checkup (July 8, v2.1.205), announced by Boris Cherny. It's an alias for the rebuilt /doctor: scans your setup and proposes one confirmable fix for (1) unused skills/MCPs/plugins costing context, (2) duplicate CLAUDE.md content, (3) an unsplit/bloated root CLAUDE.md (relocated into nested files or lazy-loaded, not deleted), (4) slow hooks, (5) a stale Claude Code version, (6) auto mode off, (7) read-only commands you keep re-approving (writes an allowlist). Edits sit in your working tree for review before commit. WHY IT MATTERS: every skill description + MCP tool def sits in context every turn whether triggered or not ("context tax"); Anthropic's playbook flags cost past 8–12 skills. One dev's 30-day audit: 31 skills → 8, faster + cheaper.
The move: update Claude Code, run /checkup in your heaviest repo, read the plan, approve. Commit the allowlist so teammates inherit it.
Cowork connects messaging/tasks/dashboards in parallel and assembles a single briefing; guided setup optionally uses your role to personalize, connects email/calendar/docs, defaults to 8:00 AM local, configurable. Set-up (no code): open Cowork, type /schedule, answer 3 questions. Since Cowork moved to the cloud (July 7), scheduled tasks run on Anthropic's servers even with your device offline (leave the folder empty / pick a cloud project). CAUTION: one complex Cowork task ≈ dozens of chat messages of quota — on Pro, keep it to one brief.
The move: schedule one weekday 8am brief tonight; open a sorted list tomorrow instead of a raw inbox.
Fable 5 extended again to July 19 (callback, no new delta)
Anthropic extended included Fable 5 access on all paid plans "at no extra cost through July 19" — the second extension in a week. Exhaust it and you can still buy usage credits. Plan around July 19, not a hoped-for permanent return.
Your context window is a budget, and you've been overdrawn.
Level up
Update Claude Code and run /checkup in your heaviest repo — read the plan, delete the skills/MCPs you never trigger. Then schedule one Cowork Morning Brief for 8am tomorrow. ~10 minutes; a lighter context bill on every message after.
The written brief from each episode, free, every morning.
Transcript
0:00Nova: Claude Code just shipped a one-command cleanup that hunts down every skill, every MCP server, every bloated CLAUDE dot M D line that has been quietly taxing your context on every single turn, and then it fixes them for you. It is called slash checkup, and most builders have not run it yet. This is Mainframe, your daily guide through the AI chaos: what actually happened, who is winning, and how to take yourself to the next level. It is Monday, July thirteenth, twenty twenty-six. I am Nova.
0:37Nova: On today's Mainframe: the slash checkup command that audits your Claude Code setup and claws back the tokens you have been leaking every message. Anthropic ships a guided Morning Brief inside Cowork, a scheduled agent that sorts your day before you open your laptop. Claude just got cheaper (or more expensive) depending on where you live, because rupee pricing landed in India this morning, and there is a wallet lesson in it for everyone. And a quick callback on Fable 5, which lives another week. Let's get into it.
1:13Dex: Quick note before we go: the voices you are hearing are AI. The reporting, the picks, and the opinions are put together by humans. Now, the Board.
1:24Nova: One macro move that actually touches your wallet today: Anthropic started localizing Claude pricing outside the United States. Anthropic has begun localizing Claude's pricing in India, its biggest market outside the U.S., and local pricing has started to appear for some users on Claude's website and mobile apps. Claude Pro now starts at one thousand nine hundred ninety-nine rupees per month, roughly twenty-one dollars. Here is the builder read: this is not charity, it is a defensive price war. The twenty-x Max tier can cost up to twenty-three thousand nine hundred ninety-nine rupees per month, and all paid Claude plans in India now include tax.
2:12Nova: Competitors like OpenAI and Google already introduced India-specific pricing with localized payment options. When the labs start fighting on local currency and taxes, that pressure eventually reaches your API bill too. So the discipline is the same as it has been all month: do not lock your pricing to one vendor's sticker price, because the floor keeps moving.
2:39Dex: Which is exactly the world our sponsor was built for.
2:46Nova: Brought to you by Outpace. Picture an operator who spent thirty years actually building things: real estate, hospitality, food, even a theater. Now that same operator points a whole team of AI agents at one client's project at a time, strategy through ship. Outpace is a system, not a freelancer, and not a faceless mega-agency. The agents do the work, one senior operator steers them, and at the end you own the code and the keys. And in a week where Claude's price literally changed by country, you want someone steering the whole fleet so you are never captive to one vendor. Book a thirty-minute call at outpace dot media. That is outpace dot media.
3:39Dex: The Wire. Start with the headliner, because this one you can run in the next five minutes.
3:45Nova: Anthropic shipped slash checkup on July eighth. Claude Code shipped a one-command cleanup: slash checkup scans your setup, finds unused skills, duplicate CLAUDE dot M D content, slow hooks, and permission prompts you keep re-approving, and proposes a fix you approve before anything changes. Boris Cherny at Anthropic announced it directly. Here is the crucial detail: it is not actually a brand-new command. Slash checkup is slash doctor's alias, not a new command. As of the v two point one point two zero five release on July eighth, slash doctor is now a full setup checkup that can diagnose and fix issues, and slash checkup is its alias.
4:35Dex: Why does this matter to a builder's wallet, specifically?
4:38Nova: Because context is not free. Every skill description, every MCP tool definition, sits in your window on every turn whether you use it or not. One dev did a thirty-day audit: started with thirty-one skills installed in Claude Code, ended with eight. The twenty-three deleted were not broken, he just was not using them enough. And this is the part that stings: each one's description still sat in the context window every turn and kept costing tokens whether it was triggered or not. Anthropic's own playbook draws the line at eight to twelve skills before the cost shows up. Past that, every line is what the playbook calls context tax.
5:29Nova: This week, update Claude Code, then in your biggest repo type slash checkup and read the plan before you approve it. Its own example describes moving, not deleting: oversized sections get relocated into nested files or marked for lazy loading, and the edits sit in your working tree for review before you commit. It also does what slash fewer-permission-prompts does: scan your transcripts and write an allowlist for common read-only commands. You will get a cheaper, faster session on every message after, and you will be ahead of everyone still carrying a thirty-skill stack they installed off a thread and forgot.
6:15Dex: Second Wire item, and it is the one non-technical operators will care about most.
6:21Nova: A guided Morning Brief inside Cowork. Tibor Blaho surfaced it first. In Cowork, Claude connects to your messaging, task tools, and dashboards at the same time, reads full conversation histories, surfaces discussions you were not part of, and pulls together what matters into a single briefing. The setup is genuinely no-code. Open Cowork, start a session, type slash schedule, Claude asks you three questions, you answer them, done.
6:53Dex: And the piece that changed last week makes this actually work unattended.
6:58Nova: Right. Since Cowork moved to the cloud on July seventh, scheduled tasks now run with no device online: set Monday's client prep for six a.m., and Claude works through the email threads, transcripts, and recent news, builds the briefing doc, and leaves the follow-up email drafted but unsent. Leave the folder empty or choose a cloud project and the task runs in the cloud, so it keeps running even when your computer is off. The move: tonight, schedule one weekday brief for eight a.m. Tomorrow you open a sorted list of what needs you, not a raw inbox.
7:43Dex: A single complex Cowork task consumes as much quota as dozens of regular chat messages. So on Pro, keep it to one brief, not ten scheduled tasks. Watch the meter.
7:59Nova: Fair. And a quick Fable callback, because we flagged the cliff. We told you Sunday the meter starts July nineteenth. That still holds. Anthropic extended access for Claude Fable 5 for subscribers for the second time in a week, saying Fable 5 would be available to subscribers at no extra cost through July nineteenth. Even if you exhaust the expanded access, you can still use the model by purchasing usage credits. No new delta beyond that: plan around the nineteenth, do not plan around a hoped-for permanent return.
8:37Dex: Tool Lab. Two picks. First one is the headliner itself, so let's give the second slot to something else.
8:45Nova: First pick, trending: slash checkup, which we just walked through, credit Boris Cherny and the Claude Code team. Update Claude Code, run it in your worst repo, approve the plan. That is the ten-minute win.
9:02Nova: Second pick, and it is a hidden gem most people missed under the Fable noise: Cursor version three point eleven shipped side chats on July tenth. Credit the Cursor team. Here is the problem it kills. The main agent thread is stateful and expensive, and every side question interrupts the agent's context and risks derailing a long-horizon task. The fix: use slash side, slash B T W, or the plus button to create a new side chat that has context from the main chat, and each side chat is a durable, full agent conversation that you can revisit later and at-mention to pull context back into the main thread.
9:51Nova: Your main agent is implementing rate limiting on an endpoint. You type slash side, ask which rate-limiting libraries work with your framework, and the side chat researches options while your main chat continues the implementation. When you like the answer, at-mention it back in. One more granular detail worth stealing: each side chat can use a different model than your main chat. So run the tangent on a cheap model, keep the expensive one on the build. It changes your day: you stop killing long runs just to ask a clarifying question.
10:28Dex: Use Cases. Clever moves with tools already in your hands. Go.
10:34Nova: One: the pre-approve audit. Most people run slash checkup once for the token savings and stop. Instead, use the permission piece deliberately. It scans your transcripts and writes an allowlist for common read-only commands. Run it, then commit that allowlist into your repo's config so every teammate and every future session inherits your pre-approved commands. You turn a one-time cleanup into a durable team default.
11:09Nova: Two: the briefing as a portable skill, not a one-off. When you set up your Cowork Morning Brief, do not just schedule it. Create a claude dot M D file in your Cowork folder, write your role, current projects, key contacts, and priorities, and Claude reads it on every run, making outputs more relevant. That file is portable: it is plain markdown, so the same context that sharpens your brief also drops into Claude Code or a cheaper model later. Write the judgment down once, reuse it everywhere.
11:52Nova: Three, a wallet move for the non-United-States listener: if you are in India, the rupee pricing means you can finally pay without a forex markup and eighteen percent tax stacked on top. Anthropic listed rupee prices directly, and these figures already include tax, so what you see is what you pay. If you were routing around dollar billing, re-price your actual plan today; the effective cost just dropped for a lot of you.
12:27Nova: Brought to you by SearchVis, an Outpace Media product. Mainframe tells you what shipped. SearchVis makes sure the models know what you shipped. When a buyer asks ChatGPT, Claude, Perplexity, or Google's AI Overviews for the best tool in your category, the model names you or it names your competitor, and if it does not name you, you do not exist to that buyer. SearchVis tracks whether the answer engines name your brand, tells you why, and gives you the exact move to publish to win the citation. Check where you stand and start free at searchvis dot outpace dot media. That is searchvis dot outpace dot media. Be the answer.
13:16Nova: Here is the through-line nobody said out loud today: your context window is a budget, and you have been overdrawn without noticing. Slash checkup exists because the whole ecosystem quietly trained us to hoard. Install every trending skill. Wire up every MCP server. Stuff CLAUDE dot M D with every rule you ever thought of. And all of it is silently billed on every single turn, whether it fires or not. The mental model to carry this week: skills and MCPs are not free features you collect, they are subscriptions your model pays for you, forever, in tokens.
14:02Nova: The counter-instinct is subtraction as a skill. The dev who cut thirty-one skills to eight did not get worse, he got faster and cheaper, because the eight that survived actually earned their place. This is the same lesson as the routing dial, the same lesson as distilling Fable into a file: leverage comes from what you own and control, not what you accumulate. A lean, audited setup is a competitive edge, because most builders will never look under the hood.
14:38Dex: So here is what you would actually do about it.
14:41Nova: So here is what I would actually do: treat your context window like a monthly bill. Once a week, run slash checkup, read the plan, and cut. Then take the one workflow you keep re-explaining and write it down as one skill or one claude dot M D file, and delete the three you never use. Cheap by default, lean on purpose.
15:08Nova: Level up this week: update Claude Code and run slash checkup in your heaviest repo, read the plan, and delete the skills and MCP servers you never trigger. Then schedule one Cowork Morning Brief for tomorrow at eight a.m. Ten minutes, a lighter context bill every message after.
15:30Dex: And if you want every story, number, and link from today in your inbox, the free Mainframe newsletter is the written version of this show. Subscribe, it is the brief we build the episode from.
15:43Nova: That is Mainframe for Monday, July thirteenth. Audit something. We will see you tomorrow.