🧭 Claudeforce: Salesforce and Anthropic Bring Claude Natively into Enterprise CRM

Salesforce and Anthropic have announced Claudeforce, an expanded strategic partnership that embeds Claude's reasoning directly inside Salesforce's enterprise data, workflows, and governance layer. The launch product, Salesforce in Claude, is a plugin with 37 prebuilt sales skills that lets sellers query live revenue context, automate pipeline updates, and take governed actions without leaving Claude.

How Salesforce in Claude works

An admin connects the Salesforce organisation to Claude once — authentication and role-based permissions are managed centrally, so every seller on the team gets access from day one with no per-user setup. Claude reads the seller's full enterprise context (accounts, opportunities, pipeline stages, Slack history, any connected CRM connector) and generates tailored interactive dashboards on first open. Skills highlighted at launch include:

• Meeting prep — pulls deal history, stakeholder map, and recent comms before a call
• Deal health review — surfaces risk signals from CRM data and Slack activity
• Pipeline review — shows stage-by-stage status with suggested next actions
• 34 additional skills covering forecasting, account planning, competitive intelligence, and renewal tracking

Governance built in, not bolted on

All actions Claude takes — creating records, updating fields, logging calls — are routed through Salesforce, not around it. Business rules, validation rules, and field-level security enforced in your Salesforce org apply automatically. This means IT and RevOps teams keep their existing governance model without writing Claude-specific guardrails.

Timeline and market response

The plugin is available to select pilot customers now, with an open beta planned for September 2026. Additional skills for service, marketing, and finance roles will launch in Q3–Q4 2026. Salesforce stock climbed 12% in after-hours trading on the announcement day, which coincided with the company's Q2 FY27 earnings release. No pricing has been disclosed; the announcement notes pricing and packaging are subject to change.

🧭 Claude Code's Linux Binary Shrinks 4.5× — CLI Now Starts Before Its Own Sandbox

The v2.1.251 release shipped a performance sprint that got less coverage than its security patches yesterday: the Linux x64 binary is now approximately 75 MB — 4.5× smaller than before — thanks to removal of six rarely-used syntax highlighting language bundles and CPU-level optimisations that also cut per-session memory by 40–70 MB. For CI pipelines and container deployments where Claude Code runs in minimal images, this substantially reduces pull times and layer cache churn.

Startup: CLI before sandbox

The most immediately noticeable change for interactive users is that the CLI's REPL now appears before the sandbox and MCP servers finish initialising. Previously, starting Claude Code in a project with many MCP servers meant staring at a loading indicator for several seconds. Now the prompt is ready immediately; MCP tools become available as each server comes online in the background. Subcommands that don't need the full runtime (version checks, config reads, --help) also skip unnecessary setup and return near-instantly.

Three new visibility features: /usage, /tasks, /permissions

• /usage per-loop breakdown — each recurring agent loop now shows its own run count and token consumption, making it easy to identify which loop is driving most of your weekly usage
• /tasks subagent model + effort display — the task list now shows which model each subagent ran on and at what effort level, so you can verify that cost-control settings are being applied as intended
• /permissions auto-mode tab — a new tab in the permissions interface lists the default auto-mode allow/deny rules in human-readable form and lets you add custom rules without manually editing settings.json

Remote Control: auto-reconnect

Remote Control sessions previously exited when the server-side disconnected (network blips, server restarts). They now auto-reconnect and resume phone notifications after reconnection. Live viewing of foreground subagent tool calls also streams through Remote Control, giving monitoring dashboards a real-time feed without polling.

# Verify you're on the updated binary after upgrading:
$ claude update
$ claude --version    # should show v2.1.251 or later
$ ls -lh $(which claude)  # confirm Linux size is ~75 MB

🧭 Claude Code's +50% Weekly Boost Expires Tomorrow at 11:59 PM PT — No Fourth Extension Announced

The promotional +50% weekly usage limit for Claude Code expires at 11:59 PM Pacific Time on August 31, 2026 — tomorrow. Anthropic has not announced a fourth extension. The boost, which has been in place since May 13, has been extended three times:

• May 13 — boost first applied; billed as a limited-time offer
• July 19 — first extension; expiry pushed from the original date
• August 19 — second extension; expiry pushed to August 31
• August 31, 11:59 PM PT — current expiry; no further extension confirmed

Who is affected

The boost applies to Pro, Max, Team, and legacy seat-based Enterprise plans. It raises weekly Claude Code limits by 50% above the published standard — for example, if your plan's standard weekly allowance is 100 sessions, you currently get 150. After expiry, you revert to 100.

The 5-hour rolling session limit is a separate mechanism and is not affected by this expiry.

What Anthropic has said

The @ClaudeDevs account stated: "We hope to make this a permanent change to our plans, but strong demand for our models means that capacity may be tight over the coming weeks." That language — "hope to make permanent" — has remained consistent across all three extension announcements without becoming a commitment. Anthropic has cited model capacity constraints as the reason it has not locked in higher limits.