1Password ARR
$500M+
Fiscal 2026 Q3, announced Oct 22, 2026; up from about $400M in Nov 2025
Net new ARR growth
~90% YoY
Fiscal 2026 Q3; about 60% across the first three quarters of the year
Non-human identity users
+500% YoY
Monthly actives on 1Password developer and AI security tools, fiscal 2026 Q3
Agent identities per human
79
Of 109 machine identities per human identity, Palo Alto Networks 2026 survey of 2,930 security leaders
The number that matters is not $500 million. It is the human tap. On October 22, 2026, 1Password crossed $500 million in annual recurring revenue and, in the same announcement cycle, shipped 1Password for Agents — a layer where an agent requests a credential, the user approves on their own device, and 1Password types the secret directly into the destination site. The company monetizes credential storage today; the product it just launched monetizes consent, one session at a time, and no market has been built to price consent yet.
Mechanism
The Broker Types the Password, So the Agent Never Sees It
This is not password rotation or a shared-secret vault for agents. It is a brokerage model with a human in the authorization path. When an agent reaches a sign-in form it presents a unique digital key so 1Password can verify which agent is asking; 1Password then asks the user, on their own device, which credential is needed and why. Once approved, 1Password fills the password and any one-time code into the site’s form itself, and the agent proceeds with an authenticated session it cannot read the secret behind. Access is granted per session, limited to the specific items approved, and is not carried over to the next task.
- The agent proves who it is with a unique digital key before 1Password will negotiate with it.
- The user approves on their own device, seeing which login is needed and for what.
- 1Password types the password and one-time code into the site’s form, keeping both outside the agent’s view.
- Grants are per session, limited to approved items, and do not carry into the next task.
- Launch partners: Muse from Meta Platforms, Claude Duet from Anthropic, ChatGPT Work and dots from OpenAI, Grok Bot, and Townie; Browserbase, Browser Use, Dana, Instinct, KERNEL, Perplexity, Ramp, Tomo and Wajo are building on it.
Principal and liability
Every Current Legal Framework Still Makes the Human the Principal
Delegated identity only works if the law agrees on who is acting, and the law has not caught up. On August 4, 2026, the Ninth Circuit vacated an injunction that had barred Perplexity’s shopping agent from Amazon, holding that because the agent acts only when a user tells it to, it is the user — not the software — accessing Amazon’s servers. That ruling opens agent commerce and closes the exit: the human remains the principal in every current doctrine, while the software does the spending. A Harvard Journal of Law and Technology digest on the same question argues the provider behind such an agent should presumptively inherit the user’s authorization unless it converts the session into its own project — an exception that has no obvious trigger.
| Party | Proposed liability when an agent overreaches | Status |
|---|---|---|
| Consumer | Not liable if the AI exceeded delegated authority | Proposed by Lead Bank, reported Oct 13, 2026 |
| Bank holding the account | Liable if it failed to control the delegated authority it granted | Proposed; no final rule |
| AI model developer | Liable for model defects or missing safety safeguards | Proposed |
| Seller or merchant | Liable for deliberate acts that induce the agent to overreach | Proposed |
| Credential broker | Not addressed in any current proposal | Holds the only approval record |
The gap is not academic. Payments firms are already writing the verification standard — a draft Agentic Financial Service Protocol proposes verifying an AI’s identity, whether the user approved the transaction, and tracking the flow for audit — while the Cloud Security Alliance argues the attribution gap is the binding constraint on AI regulation: the enterprise cannot produce a trail showing who authorized the agent. Federal Reserve Governor Christopher Waller told Sibos in September that the lack of trust, not the technology, is the biggest obstacle to agent commerce. Add Bain’s projection of $500 billion in US AI agent commerce spending by 2030 and the sequencing is clear: liability is being written faster than delegated identity is being priced.
Attack surface
The Scarce Input Is Human Attention, and Nobody Meters It
Here is the part of the launch that has not been priced. In the consumer product, every delegated login costs one human tap, and an agent that runs dozens of tasks a day will generate dozens of prompts — the same approval-fatigue weakness that makes push-based multi-factor authentication a soft control. Reporting on the launch made the parallel explicit: agents that book the wrong vacation or order the wrong item leave no option to cancel, so common sense is required, not just a technical control. That moves the scarce input in identity from stored secrets to human attention, and human attention is not metered, priced or throttled anywhere in the product.
Passkeys are winning the login and out-performing MFA on success rate
Share of successful sign-ins using a passkey, and sign-in success rates, per the FIDO Alliance Passkey Index published Oct 14, 2026.
Unit: %
Successful sign-ins using a passkey, 2026
36%
Successful sign-ins using a passkey, 2025
26%
Sign-in success rate with a passkey
93%
Sign-in success rate with traditional MFA
78%
That chart frames the credential-economics shift. As passkeys take a third of successful sign-ins, the secret itself becomes less valuable and the authorization decision becomes more valuable — which is exactly why 1Password’s own narrative moved from protecting passwords to establishing trusted access across humans, machines and AI agents. The paradox is that passkeys delete the human from the login loop while agent delegation puts the human back into it, per approval. The broker that resolves both is the asset, and the approval prompt is the meter.
Supply chain
Delegated Authority Has a Supply Chain, and the Broker Holds the Audit Trail
| Layer | Who | What moves | Evidence of linkage |
|---|---|---|---|
| Agent | Muse from Meta Platforms; dots and ChatGPT Work from OpenAI; Claude Duet from Anthropic; Grok Bot; Townie | Requests one credential for one task | 1Password launch announcement, Oct 22, 2026 |
| Credential broker | 1Password | Types password and one-time code into the site’s form; per-session, per-item scope | 1Password for Agents product documentation |
| Browser automation | Browserbase, Browser Use | Drives the page the credential is typed into | Named integrations in progress at launch |
| Enterprise control plane | Microsoft Entra Agent ID; Okta for AI agents; SailPoint | Owns agent identity, policy and audit | Microsoft 365 E7 Oct 19, 2026; Okta and SailPoint Q2 FY2027 results |
| Transaction endpoint | Amazon; Perplexity’s Comet; Ramp | Converts delegated authority into money | Ninth Circuit ruling Aug 4, 2026; partner list at launch |
Two structural points follow. First, the value sits at the layer nobody has built yet at scale: the broker is the only party that can prove a human approved a specific credential for a specific site at a specific moment — and the only party positioned to write the audit record that payment regulators are starting to ask for. Second, the volume behind it is not speculative. Palo Alto Networks’ 2026 Identity Security Landscape survey of 2,930 security leaders found 109 machine identities for every human identity, up from 82 to 1 a year earlier, with 79 of those 109 being AI agents and agent identities expected to grow 85% over the next twelve months against 77% for machine identities overall. Every one of those agents is a candidate for a delegated credential.
Pricing
The Layer Already Has Private Pricing — the Public Price Is the Part Missing
1Password’s growth is machine-shaped, not seat-shaped
Company-reported growth metrics from the fiscal 2026 third-quarter release, Oct 22, 2026.
Unit: %
Monthly actives: non-human identity users
500%
Monthly actives: human users
200%
Net new ARR growth, Q3 FY26
90%
Net new ARR growth, first three quarters FY26
60%
Reconciling the ARR figures is where the economics get interesting. ARR moved from about $400 million in November 2025 to more than $500 million in October 2026, so the stock grew roughly 25% in eleven months — yet net new ARR growth ran near 60% across the first three quarters and near 90% in the third. With gross retention at 90%, that gap implies gross additions near $140 million against roughly $40 million of annual churn: the net-new headline flatters the base, and the company is monetizing churn replacement as much as new logos.
1Password by the numbers (private company, company-reported)
ARR
$500M+ (C$702M)
Fiscal 2026 Q3, announced Oct 22, 2026
Business share of ARR
Above 75%
More than 200,000 business customers; 85% of the Forbes AI 50
Gross retention
90%
Net retention improving year over year; absolute figure not disclosed
Developer and AI tool attach
16% of business accounts
Those accounts grow ARR about five times faster
Non-human identity users
+500% YoY
Monthly actives; human users of the same tools +200%
Last confirmed private valuation
US$6.8B
January 2022 round; management says a public company is the plan
| Company | Latest reported period and revenue | Growth | Delegated-identity proof point |
|---|---|---|---|
| Okta | Q2 FY2027 (ended Jul 31, 2026): $805M | +11% | New products were about 30% of bookings; adding one lifts ACV about 40%; dozens of AI-agent deals, several above $1M |
| SailPoint | Q2 FY2027 (ended Jul 31, 2026): $309M | +17% | AI-driven ARR above $70M against a $100M fiscal-year-end target; AI adopters raised annual spend about 60% |
| Zscaler | FY2026 (ended Jul 31, 2026): $3.35B | +25% | Zero-trust access to the applications agents act on; still a GAAP operating loss |
| Palo Alto Networks | FY2026 (ended Jul 31, 2026): $11.48B | +24% | Closed its $25B CyberArk purchase on Feb 11, 2026; survey finds 79 agent identities per human |
| Microsoft | FY2026 (ended Jun 30, 2026): $331.8B | +18% | Entra Agent ID plus Dynamic Permissions in Microsoft 365 E7 cover 60 discovered agent types |
| Meta Platforms | Q3 2026 (ended Sep 30, 2026): $64.6B | +26% | Muse is the first agent in 1Password’s delegation network, rolling out across all platforms |
| Amazon | Q2 2026 (ended Jun 30, 2026): $200.6B | +20% | Ninth Circuit says an agent on a user’s instruction is the user; agent shopping can no longer be barred outright |
We believe this will be a public company some day and that is how we’re building the team and scope.
Horizons
Two Horizons: Integration Cadence Now, Identity Standards by 2027
- Integration cadence is the near-term read: Muse and Grok Bot were rolling out over the days after Oct 22, while Claude Duet, ChatGPT Work and dots were still marked coming soon.
- Geographic gating matters more than it looks — 1Password for Agents is not supported on 1Password.eu or 1Password.ca at launch, so the consumer delegation layer begins as a US-only regulatory bet.
- SailPoint’s $100 million AI-driven ARR target by its Jan 31, 2027 fiscal year end is the first dated public test of what agent governance earns.
- Okta’s next print is the tell on conversion: new products were about 30% of Q2 FY2027 bookings, yet management framed AI-agent security as an early-stage opportunity rather than a revenue driver.
- The first public price tag for delegated identity will be 1Password’s own, if the IPO it has said it is building arrives with agent-access pricing disclosed.
| Milestone or risk | Timing | What it would mean |
|---|---|---|
| 1Password lists or files publicly | Management says a public company is the plan; last mark $6.8B from 2022 | First mark on delegated identity, against ARR that has grown roughly 25% in eleven months |
| SailPoint reaches $100M AI-driven ARR | By fiscal 2027 year-end (Jan 31, 2027) | Validates a public price for governing machine and agent identities |
| Microsoft-style dynamic permissions reach consumer agents | 1–3 years | Removes the human tap and guts the approval meter 1Password depends on |
| Merchant and bank authentication standards for agents | 2027 onward | Turns approval records into a regulated artifact — and a fee |
| Approval fatigue or one agent-caused fraud case | Any time | One unrecoverable transaction can freeze consumer adoption, with no cancel path |
Falsifiers
What Would Break This View
- If consumer agent access never converts to paid tiers, 1Password’s delegation layer is a free funnel with no revenue line, and the pricing case collapses to enterprise attach rates alone.
- If dynamic permissions become standard in consumer agents, the human tap disappears and the approval meter that anchors the delegation thesis goes with it.
- If agent identity ends up bundled inside enterprise suites at no incremental cost, SailPoint and Okta lose the pricing power their AI-driven ARR lines imply.
- If approval traffic per agent stays low because agents are scoped to read-only work, the delegation layer never becomes a high-volume authorization business.
- If one incident of agent-driven fraud lands on a consumer wallet with no cancel path, merchants and banks will demand their own agent identity, and the broker moves one step further from the user.
Where delegated identity shows up in listed stocks
- New products took about 30% of Q2 FY2027 bookings and adding one lifts deal ACV roughly 40%.
- But revenue grew only 11% to $805M and management calls AI-agent security an early-stage opportunity, not a revenue driver.
- Competition: Microsoft ships agent identity inside Microsoft 365 E7, competing for the same budget line.
- Next two quarters — watch whether agent bookings convert to revenue or stay pipeline.
- AI-driven ARR passed $70 million against a $100 million fiscal-year-end target.
- Total ARR reached $1.231 billion, up 25%, and customers adopting AI-driven solutions raised annual spend about 60%.
- Cleanest public proxy for the pricing of delegated-identity governance.
- One to three years — agent counts compound with governance spend; Palo Alto Networks counts 79 agent identities per human.
- Closed its $25 billion CyberArk purchase on Feb 11, 2026, adding privileged access to agent governance.
- Its own 2026 survey finds 79 AI agent identities per human identity, with agent identities expected to grow 85% in twelve months.
- Identity is the fastest-growing line in a platform business already growing revenue 24%.
- One to three years — platform consolidation plus agent-identity growth keeps the pull-through intact.
- Microsoft 365 E7 grants agents task-scoped, time-bound access with no standing privilege.
- Entra Agent ID now covers 60 discovered agent types, including Claude Code, Codex and GitHub Copilot.
- This is the enterprise answer to the human tap 1Password’s consumer product requires: policy, not prompts.
- 2027 onward — bundling inside an existing suite is the cheapest way to own delegated identity.
- Ships Muse as the first agent in 1Password’s delegation network, rolling out across all platforms from Oct 22.
- Removing the login gate should lift completed agent tasks; Q3 2026 revenue grew 26% to $64.6 billion.
- Agent commerce distribution is where Meta’s consumer AI gets monetized, and 1Password is the on-ramp.
- One to three years — the distribution point for always-on agents becomes the merchant of record for agent transactions.
- The Ninth Circuit held on Aug 4 that Perplexity’s agent is the user, so Amazon can no longer bar agent shopping outright.
- Bull case: agent-driven buying adds incremental demand and completes checkouts that used to abandon.
- Bear case: per-session delegated grants give merchants no way to distinguish an approved agent from a frauding one.
- Next two to four quarters — merchant-side agent authentication standards decide whether the volume actually arrives.
