You can now grant Claude access to your 1Password credentials

1Password launched “1Password for Claude,” enabling Claude agents to complete real-world tasks (e.g., booking travel, managing accounts) by injecting credentials directly to target systems while preventing passwords/MFA codes from reaching the model, Anthropic systems, or agent memory. The product uses a “zero-exposure” per-session, limited-access framework with browser lock-down during agent control and post-autofill secret scanning. Availability starts on Mac for business/family/individual plans, with payment cards and identity details slated for post-launch.

Analysis

This is a trust-layer upgrade, not a headline revenue event. The important mechanism is that agentic AI can now execute authenticated workflows without the credentials ever becoming model-visible, which removes one of the last practical blockers to enterprise adoption. That shifts the debate from "can agents do tasks?" to "who owns the permissioning and audit trail?" — a much better setup for identity and security vendors than for model providers alone.

Near term, the clearest beneficiaries are the vendors that sit at the control point for delegation and session governance: OKTA for policy/identity, and PANW/CRWD for endpoint and browser containment around autonomous sessions. The second-order loser is any app layer that depends on manual logins and user attention; if agents become the interface, distribution migrates away from standalone consumer workflows and toward whoever controls authenticated access. TGT is not a clean trade here; any retail read-through is too indirect unless agentic checkout begins to show up in actual conversion data.

The market is likely overpricing the immediacy of monetization and underpricing the adoption trigger: 1-3 months from now the catalyst is management commentary, partner integrations, and evidence of enterprise pilots; 6-18 months is when payment-card and identity support could matter if this moves from demo to usage. The main falsifiers are simple: no meaningful uptake beyond early adopters, or native OS/browser vendors (MSFT/GOOG/AAPL) absorbing the permissioning layer and commoditizing the middleware. A single credential leak would also slow the whole category by forcing a tighter review of agent permissions.

Key Decisions for Investors

• No direct trade in TGT; treat this as a sector read-through only. Reassess only if retail/commerce data starts showing agent-driven checkout lift or if TGT discusses AI-assisted account actions on earnings.
• Tactical long OKTA on any post-news weakness, 1-3 month horizon. The setup is that agentic identity becomes a budget line item; invalidate if management calls out no measurable demand pull from delegated-auth use cases.
• Add PANW/CRWD to the watch list rather than chase immediately. Buy only if enterprise commentary shows demand for agent containment, audit, or browser/session controls; otherwise this remains a product-demo story, not a P&L story.
• Use MSFT and GOOG as competitive risk hedges against the middleware stack. If browser/OS-native permissioning arrives faster than expected, it compresses the valuation of standalone agent-security layers over 6-18 months.
• Set an alert for any reported credential incident involving an AI agent ecosystem. That would be the fastest way to repriced the theme lower and would likely unwind the current optimism in identity/security proxies.