Claude Code weekly limits go up 50% through July 13, and the week in Claude
PLUS: Claude for Small Business ships with 15 workflows, eligible users can claim an Agent SDK credit from June 15, and Claude Code adds agent view in research preview
Morning, builders. Here's your week in Claude, from May 10 to May 16.
This week, Claude Code weekly limits went up 50% for now. Claude for Small Business arrived with 15 workflows. And eligible users can claim a monthly Agent SDK credit from June 15.
Plus 10 quick hits. Every claim has a link.
Claude Code weekly limits go up 50% through July 13
Pro, Max, Team and seat-based Enterprise users get 50% more weekly Claude Code use, for now.
On May 13, Anthropic's ClaudeDevs account posted the news. It's live for all of those users.
Per a follow-up post, it runs through July 13 at 6 pm PDT. It applies everywhere you use Claude Code. That's the CLI, IDE extensions, desktop and the web.
There's nothing to opt into. It stacks with the 2x increase to 5-hour limits announced the week before.
Claude for Small Business puts Claude inside QuickBooks, PayPal and more
Anthropic launched connectors and ready-made workflows for small business owners.
Per Anthropic's May 13 post, you turn it on inside Claude Cowork. It works with Intuit QuickBooks, PayPal, HubSpot, Canva and Docusign. It also works with Google Workspace and Microsoft 365.
It ships with 15 workflows and 15 skills. They span finance, operations, sales, marketing, HR and customer service. Examples include planning payroll, closing the month and chasing invoices.
Claude does the work. You approve before anything sends, posts or pays. Per Anthropic, you can approve the plan first or, when you're ready, let it run end to end.
Anthropic and PayPal also made a free online course for owners. A free workshop tour starts May 14 in Chicago, for 100 local small business leaders per stop. Its spring stops cover 10 US cities.
Eligible plans can claim a monthly Agent SDK credit from June 15
Anthropic says Agent SDK use will stop counting against your plan's limits on June 15. Eligible users can claim a separate monthly credit for it.
That's per Anthropic's help article (archived May 13). The credit covers the Agent SDK and claude -p.
It also covers Claude Code GitHub Actions. Third-party apps that sign in with your plan through the Agent SDK count too.
The credit runs from $20 a month on Pro to $200 on Max 20x. Standard seats on seat-based Enterprise plans can't claim it. API key accounts get no credit and keep paying as they go.
Credits are per user and don't roll over. You claim yours once. It then refreshes each billing cycle.
When one runs out, use moves to extra usage at standard API rates. That only happens if you've turned extra usage on. If not, Agent SDK requests stop until the credit refreshes.
Anthropic says your plan's limits stay the same, for chat, Cowork and interactive Claude Code.
Claude Code adds agent view, in research preview
Claude Code got one screen for every session. It also got a command that keeps Claude working toward a goal.
Per its weekly digest, agent view opens with the claude agents command. It shows what's running, what's waiting on you and what's done. Background sessions keep running without a terminal.
Agent view is a research preview. Anthropic's May 11 post says usual rate limits apply.
The new /goal command lets you set a finish line. After each turn, a fast model checks if the goal is met. If not, Claude starts another turn.
Fast mode, also a research preview, now runs on Opus 4.7 by default, not Opus 4.6. The digest says it's the same model quality at about 2.5x the speed, for a higher per-token cost. Its price stays $30/$150 per million tokens (input/output).
Claude for Legal adds 20+ connectors and 12 plugins
Anthropic added more tools for lawyers and legal teams.
Per Anthropic's May 12 post, the 20+ new MCP connectors link Claude to legal software. They include Thomson Reuters CoCounsel, Harvey, Docusign, Box and iManage.
Each of the 12 new plugins is built around a legal role. Examples are commercial, employment, privacy and litigation work. Anthropic says legal professionals are now the most engaged Cowork users of any knowledge-work function.
Qualifying legal aid clinics, public defenders and nonprofit legal groups can get significantly discounted pricing. That's through Claude for Nonprofits. Per TechCrunch, Anthropic made the new tools available to all paying Claude customers.
Anthropic and the Gates Foundation commit $200 million
Anthropic is teaming up with the Gates Foundation on global health, life sciences, education and economic mobility.
Per Anthropic's May 14 post, the $200 million covers grants, Claude credits and technical support. It runs over the next four years.
The largest part focuses on health in low- and middle-income countries. Anthropic says its work on neglected diseases starts with polio, HPV and eclampsia/preeclampsia.
The deal also covers K-12 education tools in the US, sub-Saharan Africa and India. Anthropic says the first public goods from that work, like benchmarks and datasets, come out later this year.
AWS customers can now get the full Claude Platform
Claude Platform on AWS is now generally available.
Per AWS's May 11 post, Anthropic runs the service. Customer data is processed outside the AWS security boundary.
Sign-in runs through AWS IAM. Billing comes on your AWS invoice. Per Anthropic, new features and betas ship there the same day they reach the Claude API.
Claude is still on Amazon Bedrock, where AWS is the data processor. Anthropic says Bedrock fits firms with strict regional data residency needs.
Quick hits
- PwC will roll out Claude Code and Cowork (May 14). Per Anthropic, it starts with US teams and expands toward hundreds of thousands of staff. The firms will also train and certify 30,000 PwC staff on Claude.
- Fast mode on the Claude API now supports Opus 4.7 (May 12). It's a research preview with a waitlist. Pricing matches Opus 4.6 fast mode.
- Cache diagnostics is in beta on the Claude API (May 13). Per the release notes, when you ask for it on a request, it reports where your prompt cache broke from the previous turn.
- Claude.ai had elevated errors on May 13, from 12:21 to 14:45 UTC (7:21 to 9:45 am Central). Per the status page, it hit claude.ai and Claude Code. An earlier incident ran from 11:25 to 11:57 UTC.
- Ramp data shows more businesses paying for Anthropic than OpenAI (May 13). Per TechCrunch, it's 34.4% versus 32.3%. The index only covers firms that use Ramp.
- Anthropic laid out two scenarios for US-China AI in 2028 (May 14). Per its paper, democracies can get a commanding lead if policymakers act now. It lists tighter chip controls on Chinese labs, disrupting distillation and faster AI adoption in democracies.
- Claude Code 2.1.141 adds "Summarize up to here" to the Rewind menu (May 13). It shrinks earlier context and keeps recent turns.
- Claude Code 2.1.143 stopped falling back to rm -rf when worktree cleanup fails (May 15). That prevents loss of gitignored or in-progress files.
- Claude Code 2.1.143 stops stop hooks from looping forever (May 15). The turn now ends with a warning after 8 blocks in a row. A setting can change that cap.
- Claude Code 2.1.142 fixed a 60-second cap on remote MCP tool calls (May 14). The MCP_TOOL_TIMEOUT setting now works for remote HTTP and SSE servers.
Worth your time
- How Claude Code works in large codebases (May 14): patterns from teams using Claude Code in huge monorepos and old legacy systems.
- The founder's playbook (May 14): Anthropic's guide to building an AI-native startup, from idea to scale.
- Best practices for computer and browser use (May 13): Anthropic's notes on screen scaling, clicks and thinking effort for computer use.
That's the week. The next issue lands Sunday, May 24.
Reply and tell me what you're building. I read every reply.
— Sam