Five of the 14 client portal examples below let a client in with an emailed link or a one-time code and never ask for a password, and only three publish a native client app with its own passcode or biometric login (MyChart, TaxDome and Canopy, per their own pages, read 9 November 2026). The rest still run on invitations, passwords and a web page that happens to fit a phone.
This page scores 14 portals that real firms put in front of clients: healthcare, legal and accounting portals, agency portals, support portals, a billing portal and a shopper account. Each is scored 1 to 3 on onboarding, permissions and mobile, from one vendor page per product that we opened on 9 November 2026. Where a vendor's help centre blocked us, we say so and score what we could read.
How we scored the client portal examples
We read one page per portal, the page a buyer or a client would be sent to, and scored only what that page says. The scores are ours, not the vendors'. A low score can mean a thin page rather than a thin product, and we flag the three cases where that is likely.
| Score | Onboarding | Permissions | Mobile |
|---|---|---|---|
| 3 | Link or code sign-in documented with its rules (expiry, one use, device memory) | Per-item or per-contact controls with the default stated | Native client app with a passcode or biometric login documented |
| 2 | Invitation plus password, with a second factor or SSO documented | Per-organisation, per-folder or role-based controls with a stated default | Native client app named, login not described |
| 1 | Sign-in not described on the page we read | Controls only named, or a single user who sees only their own data | Web only, or mobile not addressed on the page we read |
Source: Octo Agency rubric, applied 2026-11-09. Raw scores are in our research file for this post.
The rubric rewards documentation that a client could follow and a buyer could audit. It does not reward feature lists. "Role-based access" on a marketing page earns a 1 because it tells a buyer nothing about who sees what by default. The question a buyer should be asking of any portal is the one our client portal requirements checklist spends seven of its 40 questions on: what does a new client see on day one, before anyone has changed a setting?
The 14 portals at a glance
| Portal | Who logs in | How a client gets in | Permission unit | Client app | Score of 9 |
|---|---|---|---|---|---|
| Canopy Client Portal | Accounting clients | Email invite, password, emailed code, 30-day device memory | Per contact and per folder | Yes, passcode and biometrics | 8 |
| MyChart (Epic) | Patients | One Epic ID across organisations; passkey or biometric on the app | Proxy access per family member | Yes, passcode, biometric or passkey | 7 |
| Clio for Clients | Law firm clients | Emailed sign-in link, one use, expires in 1 hour | Per matter portal | Yes, push notifications | 7 |
| Assembly | Agency and firm clients | Google or magic link, password optional | Per contact, company and app | No app described | 7 |
| Moxo | Clients and vendors | One secure link, no account or password; SAML SSO | Role-based, separate portals per audience | Not on the page read | 6 |
| HubSpot support portal | Support customers | Access-group registration email or self-registration | Own, company or filtered company tickets | Not mentioned | 6 |
| Shopify customer accounts | Shoppers | Six-digit email code, Shop passkeys, Google or Facebook | Own orders and profile | Shop app sign-in | 6 |
| Salesforce Experience Cloud | Enterprise customers | Optional self-registration with email verification | Private by default, explicit sharing rules | Not mentioned | 6 |
| Zendesk help center | Support customers | Optional sign-in, SSO, verified email addresses | Own requests or shared organisation | Not mentioned | 5 |
| Jira Service Management | Support customers | Invited only or open; account creation is global | Organisation, typed email or whole project | Not mentioned | 5 |
| Stripe customer portal | Subscribers | Email address, then an emailed login link | Single customer, own billing | Hosted web page | 5 |
| Basecamp client access | Agency clients | Added by name or from a company on the Clients tab | Per item, private by default | Not mentioned | 5 |
| TaxDome Client Portal | Accounting clients | Face ID, fingerprint or 2FA on the app; invite not on the page read | Switch between several accounts | Yes, biometric or 2FA | 5 |
| ShareFile for Clients | Professional services clients | Password with MFA; activation not on the page read | Per folder and per file | Yes, iOS and Android | 5 |
Source: one vendor page per portal, listed under Sources; Octo Agency scoring, 2026-11-09
Three scores need a caveat before anyone quotes them. TaxDome's help centre and marketing site returned a bot wall to our requests, so its row is scored from the App Store listing alone, which says nothing about the invitation flow. Moxo's support articles were unreachable too, so its mobile score comes from a marketing page that only says "any device". Basecamp's "Working with Clients" article is about visibility, not sign-in, so it scores a 1 on onboarding without that meaning Basecamp's sign-in is weak. The other eleven rows are scored from help-centre or documentation pages.
Onboarding: five portals have dropped the password
The clearest split in the set is between portals that treat a client's email inbox as the credential and portals that still issue a password.
| Pattern | Portals | What the documentation states | Trade-off |
|---|---|---|---|
| Emailed link, no password | Clio, Stripe, Assembly, Moxo | Clio: each link works once and expires after 1 hour. Stripe: the customer enters an email and Stripe emails a login link; the email cannot be changed through that link | Nothing to forget; the inbox becomes the account |
| One-time code, no password | Shopify, Canopy (as a second step) | Shopify: a one-time six-digit code, sessions up to 365 days. Canopy: a "Client Portal Login Verification" email, with a 30-day "remember this device" option | Fast on phones; a shared inbox is a shared login |
| Invitation plus password | Canopy, TaxDome | Canopy: create account from the invite, name, password, then the emailed code | Familiar; adds a reset flow to support |
| Single identity across providers | MyChart, Shopify (Shop) | MyChart: one Epic ID for every participating organisation, with passkey, biometric or four-digit passcode on the app | One login to protect; one login to lose |
| Registration gated by the business | HubSpot, Jira, Zendesk, Salesforce | HubSpot: access-group membership with a registration email, or self-registration. Jira: "Allow customers to create accounts" is a global setting, not per project | Control over who registers; more admin work per client |
Source: Clio, Stripe, Assembly, Moxo, Shopify, Canopy, MyChart, HubSpot, Atlassian, Zendesk and Salesforce pages, read 2026-11-09
The link is the account. Clio's and Stripe's pages both make the same quiet point: when sign-in is an emailed link, the firm controls which address receives it. Clio says Firm Admins can change the client's portal email and that other users see the field read-only. Stripe says that if several customers share one email, the portal picks the most recently created customer with an active subscription. Both are permission decisions dressed as onboarding details.
Sessions are long on purpose. Shopify documents sessions of up to 365 days and a sign-in that is not re-challenged for each action. Canopy's 30-day device memory and MyChart's passcode do the same job: make the second and tenth visit cheap. For a client portal that people open four times a year, the re-entry path matters more than the first one.
Self-registration is a liability until it is needed. Salesforce Security's March 2026 guidance for Experience Cloud is the most direct text in the set: if the site does not require visitors to create their own accounts, disable self-registration; if it does, run the handler with sharing, assign the most restrictive profile and require email verification before activation. Jira's documentation echoes it from the other side: account creation is a site-wide switch, so a firm that wants an invite-only portal for one project has to restrict domains for all of them or run a separate site.
Permissions: per item, per organisation or per record
The portals split into five families, and the family decides how much thinking a client has to do.
| Model | Portals | Documented default | Where it breaks |
|---|---|---|---|
| Per item or per contact | Basecamp, Canopy, Assembly | Basecamp: "Everything in a client project is private to your team by default", with visibility badges per item. Canopy: contact-level permissions and folder-level visibility replaced "the previous all-or-nothing sharing model" in November 2025 | Staff forget to flip the badge; the client sees nothing and emails instead |
| Per organisation or company | HubSpot, Zendesk, Jira | HubSpot: own tickets, all primary-company tickets, or company tickets only for contacts meeting property criteria. Zendesk: own requests by default; a shared organisation shows all members' tickets. Jira: share with own organisation only, a typed email, or anyone in the project | A contractor attached to the wrong company sees the whole company's history |
| Per container, all-or-nothing inside | Clio, ShareFile, Moxo | Clio: participants have access to everything shared in the portal, including items shared before they joined. ShareFile: what a client can do with a file depends on the folder permissions the provider set | One over-shared matter or folder cannot be partially undone |
| Record-level sharing rules | Salesforce, MyChart | Salesforce: Default External Access Private; no record reaches an external user unless a sharing rule grants it. MyChart: proxy access per family member, plus a log of who accessed the record | Correct and expensive; every new object needs a rule |
| Single customer, own data | Stripe, Shopify, TaxDome (as read) | Stripe: one customer sees and edits their own billing; the business configures which changes are allowed. TaxDome: switching between several accounts, for couples or multiple businesses | No shared view for a team of two at the client |
Source: Basecamp, Canopy, Assembly, HubSpot, Zendesk, Atlassian, Clio, ShareFile, Moxo, Salesforce, MyChart, Stripe, Shopify and TaxDome pages, read 2026-11-09
Private by default is the feature. Basecamp's and Salesforce's pages, aimed at very different buyers, say the same thing: nothing is visible until someone makes it visible. Basecamp adds that removing a client company reverts its items to private, and that a removed client loses access immediately without being notified. Salesforce adds a rule that applies to every portal on this list: restrict portal and site user visibility so that an external user cannot enumerate the internal members.
Company-level sharing needs a property filter. HubSpot's third option, company tickets only for contacts who meet defined property criteria, is the pattern that makes organisation-level sharing safe. It lets a firm give the client's finance lead every invoice conversation without giving the intern the same view. Zendesk's end-user page puts the same control in the admin's hands: a member of a shared organisation sees every member's tickets, and can comment on them only if an administrator enabled it.
Clio names the hard case. Its warning that a re-added participant regains access to everything shared while they were removed is the clearest statement in the set of what per-container permissions cost. The fix the page offers, remove the participant, delete the message or stop sharing the document, is manual. A buyer evaluating any matter-style or project-style portal should ask whether the same is true there.
Mobile: three portals treat the app as the product
The accounting and healthcare portals are the only ones that document a phone login step by step. Canopy's first-login flow is the fullest: download "Canopy Client Portal", create the account from the invite, enter the emailed code, choose which of nine notification types to receive, set a four-digit passcode used at every login, and optionally turn on biometrics. Its November 2025 notes add that the same per-contact file permissions apply on mobile as on the web. MyChart lists four quick-login options on its app, fingerprint, face recognition, a four-digit passcode or a passkey, and location-based check-in at the clinic. TaxDome's listing promises Face ID, fingerprint or 2FA depending on device, and account switching for couples and multiple businesses.
Three more name an app without the login detail: Clio for Clients sends push notifications for portal activity, ShareFile publishes ShareFile for Clients on iOS and Android, and Shopify's sign-in with Shop uses saved passkeys. The remaining eight pages, Assembly, Moxo, Basecamp, HubSpot, Zendesk, Jira, Stripe and Salesforce, describe no app and no phone login; Moxo's says only that the link opens on any device, and Stripe's portal is a hosted page on Stripe's own domain.
The pattern is not an accident. The portals with documented apps are the ones where the client's job is to send things: receipts, scanned documents, a photo of a form. TaxDome lists scanning receipts and completing organisers. Where the client's job is to read and approve, a responsive web page is what the vendors ship, and the documentation says nothing because there is nothing to configure. Our comparison of 12 client portal products found the same gap on pricing pages: branding and permissions get a column, mobile gets a footnote.
What the examples do not show
Four things are missing from every page we scored, and a buyer should ask for them directly.
Usage. Not one page says what share of invited clients ever log in. The adoption data that does exist is sector-level, and our client portal adoption statistics post collects it; it should temper any plan that assumes a portal replaces email on day one.
The first-login failure path. Even Canopy's login article, the fullest in the set, stops at "enter the code". What happens when the link expired, the inbox is shared, or the client's IT blocks the sender is undocumented everywhere.
Audit trails for clients. MyChart's "see who has accessed your record" is the only client-facing access log in the set. Moxo and Assembly list audit logs as features, for the firm, and Assembly's sits on its $499 a month Advanced plan with enforced MFA.
What the page is for. Five of the 14 pages are a marketing page, a blog post or an app listing rather than documentation. Where we could choose, we chose the help centre, and it did not buy a higher score: the nine documentation rows average 5.8 out of 9, the five others 6.2. It bought something else, a row we can defend line by line. A marketing page that says "role-based access" scores a point; a help article that says what a new client sees by default earns the three.
What this means for your own portal
If you are choosing a product, use the three columns as a demo script. Ask the vendor to show a client's first login on a phone, to show what a brand-new client sees before anyone changes a setting, and to show how a shared item is un-shared. The 14 examples above suggest that the first and third questions are where demos go quiet.
If you are building, the examples set the bar for what "done" means. We build applications in Next.js, and on a portal we would write the permission defaults before the first screen: everything private, sharing per item or per contact, a visibility flag the client never sees, and a removal path that revokes access the moment it runs. We would treat sign-in as an emailed link with a stated expiry rather than a password, and treat the phone as the primary device for anything the client has to send. Whether a product or a build is the right answer for your situation is the subject of our buy, build or Wized decision; our web apps service page explains how we scope a portal when the answer is build.


