Rolodex

The home page headline promises artificial intelligence and automation unleashing relationships at scale. The pricing page, which is the vendor's own enumeration of what a buyer actually receives, lists map view, filters and lists, network sync, import from a customer relationship system, and priority support. Not one model driven capability appears on it.

The named product surfaces are mostly deterministic. Syncing mailbox and calendar, matching correspondence to a contact record, consolidating touchpoints at company level, tracking a job change and firing a reminder are integration and rules work. They are useful and they are not inference. Suggesting a warm introduction path is the one function where a model could plausibly do real work, and nothing states that one does.

The genuinely interesting artefact inverts the usual arrangement. A model context protocol server exposes the customer's relationship data to whatever AI client the customer already runs, rather than the vendor running a model over that data and selling the output. That is a defensible architecture and arguably a more honest one for a company this size, since it puts the intelligence where the buyer has already paid for it. But it means the model sits in somebody else's product, and the vendor supplies context rather than cognition.

So the headline claims a model and the price list does not contain one, while the real capability delegates the model to the buyer. Ask which functions run a model and which are rules.

Autonomy is deliberately low and the vendor does not pretend otherwise. The system ingests, organises, surfaces and reminds; a person decides and acts. Nothing sends on the buyer's behalf, nothing contacts anyone, and job change alerts and keep in touch prompts are notifications rather than actions. For this product class that restraint is correct, and it is why the axis does not turn on runaway automation.

The control that does exist is consent to connection, and it is real. Account connections are opt in, the security page states that the user decides whether an account is connected at all, permissions are revocable at any time, and each model context protocol client is scoped to one chosen workspace with access withdrawable from settings. Those are meaningful limits on what the system can reach.

The unaddressed layer is workspace governance, and in a product whose entire premise is pooling private networks into a shared asset, that is where the real question sits. Nothing describes whether an administrator can compel or block a member's connection, how much of a colleague's synced correspondence another member can read, whether notes about a contact are visible team wide by default, or what becomes of a departing employee's synced contacts and history. Members are contributing their own mailbox and network into a shared store with no published rules about who sees what.

Ask what a member can see of a colleague's synced correspondence, and what happens to their contacts when they leave.

Artificial intelligence appears in the home page headline and in no feature description anywhere on the site. Two passes located no model, no provider, no architecture, no version and no evaluation of any kind. There is nothing to disclose about because nothing model driven is specified, which is a different failure from concealment and lands in the same place for a buyer.

The transparency question that genuinely matters here runs in the opposite direction to the usual one, and the vendor does not engage it. The model context protocol server hands contacts, companies, emails, meetings and notes to an external AI client chosen by the customer. That raises three questions no retrieved page answers: what the server actually exposes and at what granularity, whether a full mailbox history can be pulled through it or only contact records, and whose terms then govern the data once a third party model provider has received it. The vendor documents how to connect a client in four steps and does not document what the client can then see.

For a buyer this is the sharper disclosure gap, because connecting a general purpose assistant to a store of correspondence with customers, investors and partners moves that material into a system with its own retention and training terms. The vendor is best placed to state the boundary and does not.

Ask exactly which fields and record types the server exposes, whether retrieval is logged, and what the vendor's position is on external model providers retaining retrieved content.

Two passes across the home page, pricing, security, the model context protocol page and the documentation guide located no customer outcome claim of any kind. No case study, no named reference customer, no logo wall, no metric, no testimonial and no quantified result appears on any vendor surface.

The absence is more pointed here than it would be on a general purpose tool, because the product's premise generates an obvious and easy measurement. A platform sold on surfacing warm introductions and preventing duplicated outreach could publish how many introduction paths it surfaced, how many job changes it caught before the team noticed, or how often it prevented two people contacting the same prospect. Those numbers exist inside the product by construction, since the system tracks exactly these events in order to display them, and none is published.

What third party material exists is thin and not independent. A review platform profile describes users across sales, venture capital, consulting and executive teams and states that the product has grown steadily since launch, but that text is supplied by the seller rather than derived from reviews. No review volume, analyst coverage, benchmark or independent evaluation was located.

Ask for two reference customers of your own size and motion, and for the count of introduction paths surfaced per active user per month.

Most of this axis is not applicable in the provider sense and is rated accordingly rather than penalised. The platform sends nothing. There is no sequencer, no bulk mail, no dialer and no messaging channel, so consent capture, suppression, calling hours and unsubscribe handling all belong to whatever tool the buyer actually uses to reach people. The vendor makes no claim otherwise and does not decorate itself with compliance language it has no need of.

What remains inside the record is ingestion rather than transmission, and it is not trivial. Syncing mailboxes and calendars means the vendor processes correspondence with third parties who never agreed to anything and were never told a record exists. Pooling that history into a shared workspace compounds it, because a colleague can then read the interaction history of an exchange they were not party to. The people in the database are the ones with an interest here, and they are not the customer.

Nothing published states a lawful basis for processing third party correspondence and contact data, and two passes located no European data protection material, no data processing agreement, no retention period for ingested correspondence and no route for a contact to object or be removed. The security page is written entirely for the account holder.

Ask on what basis third party correspondence data is processed, how long it is retained, and what happens if a contact objects to being in the workspace.

The commitments made are specific and unusually clean for a company this size. Data received from connected third party sources is never given to third parties, never sold and never used for advertising, stated in three explicit negatives rather than one vague assurance. Account connections are opt in and permissions revocable at any time. Deletion of an account or workspace is stated to erase all records permanently.

The Google limited use undertaking is the strongest element and it is properly formed: a named policy, linked, covering data obtained through those interfaces, including the limited use requirements by name. That is an enforceable obligation with a real scope boundary rather than a self description, and it governs the most sensitive ingest path this product has, which is the mailbox.

Two gaps hold it mid band. Everything published is written for the account holder, and this product's database is largely composed of people who are not account holders. Their contact records, correspondence history, meeting history and the notes team members write about them carry no described retention period, no rights route and no notice.

And no regional privacy framework is named anywhere across two passes, for either party. For a product processing mailbox contents there is no European material, no Californian section and no transfer position.

Ask for the retention period on ingested correspondence and the process for removing a third party contact on request.

Provenance is clean and it should be said plainly, because it is rare in this lane. Records arrive from the buyer's own accounts: professional network connections, mailbox, calendar, and an import from an existing customer relationship system. Two passes located no purchased database, no enrichment waterfall, no contact marketplace, no aggregated third party source and no vendor supplied records at all. A buyer is not inheriting anyone else's licence terms, indemnity exposure or consent problem, which removes the single largest liability in the contact data category.

The exposure is different in kind rather than absent, and it is created by the product's own premise. An individual's mailbox contacts and correspondence history are ordinarily held by that individual for their own purposes. Surfacing the same records into a shared workspace so that colleagues can see them is a change of purpose and a change of controller, and nothing published addresses it. The question sharpens on departure: nothing states whether a leaving member's synced contacts and correspondence history remain in the workspace, are removed with them, or are retained by the employer as a company asset. In a product explicitly sold on converting individual networks into a durable team asset, that is the provenance question that will eventually be litigated somewhere.

The absence of any freshness or accuracy claim is worth noting too, though the records are the buyer's own.

Ask who owns synced contacts on departure, and whether the workspace retains a member's correspondence history after their access ends.

This is the material exposure on the record and it is structural rather than incidental.

Professional network synchronisation is a headline capability. It is named on the pricing page as one of five things the plan includes, and described on the home page as automatically syncing connections and tracking job changes in real time. The vendor also publishes a browser extension for the product family.

That platform does not offer general connection export or continuous job change monitoring to third parties as a licensed capability, and its terms restrict automated collection of member data. Two passes located nothing on any vendor surface naming a partner programme, a sanctioned interface, an authorised data path or any mechanism at all for how the sync is performed. The capability is sold and the basis for it is undescribed, and for this axis the description is the thing being graded.

The consequence falls on the buyer rather than the vendor, which is what makes it worth flagging rather than merely noting. The connection is made with the customer's own account, so any enforcement lands on the customer's account and the accounts of their team members, not on the vendor's infrastructure. A sales team that has built its shared network on this basis carries that risk personally.

The mailbox and calendar side is the opposite case and is credited: the Google limited use undertaking shows the sanctioned route being used where one exists.

Ask in writing how network synchronisation and job change tracking are performed, under which programme, and who bears the consequence of an enforcement action.

The security page describes a coherent programme at a level proportionate to the company: encryption in transit and at rest, workspace access policies limiting who can view or manage data, continuous monitoring, vulnerability scanning and periodic audits. The data commitments are the strongest part and are stated as absolutes rather than intentions: integration data is never given to third parties, never sold, never used for advertising, and can be removed or its permissions revoked at will.

The stewardship question left open is training, and the model context protocol server converts it from theoretical to immediate. A server that hands contacts, companies, emails, meetings and notes to an external assistant raises two distinct questions, and the vendor answers neither. Whether the vendor trains on workspace content is squarely within its control and unstated. What happens to material once a third party model provider has received it is largely outside its control, which is precisely why a buyer needs the vendor to draw the boundary explicitly rather than leave it implied.

Claims also diverge across pages. The model context protocol page promises enterprise grade security, role based access and full audit logs. The security page describes user access policies and names no audit log capability, no role model and nothing that would support the enterprise grade characterisation. One of the two pages is overstating, and a buyer relying on the audit log claim should confirm it exists before assuming it does.

Ask whether workspace content trains any model, and confirm the audit log capability is real before relying on it.

The transmission half of this axis is not applicable in the provider sense and is rated accordingly rather than penalised. The platform sends no message, generates no outbound copy, synthesises no voice and impersonates nobody. There is no recipient of anything the vendor produces, so sender authenticity and automation disclosure do not arise.

The half that does apply concerns the people held in the database rather than recipients of outreach. This product builds a durable and detailed record about third parties: their contact details, the full history of correspondence with anyone on the team, meetings attended, notes colleagues have written about them, and their employment tracked over time. None of them is told the record exists, and no notice, objection route or visibility mechanism appears on any retrieved surface.

That is ordinary for the contact management category rather than a failing unique to this vendor, and it is the reason this sits mid band rather than lower. Every shared address book carries the same property.

One capability does push past the category norm and deserves naming. Real time job change tracking means the platform continuously monitors named individuals' employment status and alerts the team when it changes. Monitoring a person's career on an ongoing basis, so that a team can time an approach to the moment they move, is closer to surveillance of a named individual than a static address book is, and the vendor markets exactly that timing benefit.

Ask what a contact would see if they asked what the workspace holds about them.

Integration is the product rather than a bolt on, and the ingest surfaces are named specifically: professional network, two mailbox providers, calendar, and an import path from an existing customer relationship system. Native applications ship on both mobile stores and a browser extension is published, so the product reaches the places a person actually keeps their contacts.

The distinguishing element is a model context protocol server, and it is shipped rather than announced. The endpoint is published in full, documentation lives at its own subdomain, authentication is by account sign in with no key management, each connecting client is scoped to a chosen workspace, and access is revocable from settings. Named clients include the major assistants and a command line tool. A working endpoint with an authorisation flow, per workspace scoping and revocation is a real artefact that a buyer can test in minutes, and shipping one at this point in the cycle puts this vendor ahead of most of the lane on the interface that is becoming the relevant one.

The gap is direction. Everything named moves data inward. Two passes located no conventional application interface, no webhooks, no automation platform connector and no write back path to a system of record. Import from a customer relationship system is offered and export to one is not, so a team running this alongside a pipeline tool must maintain the connection manually.

Ask whether anything writes back to your system of record, and confirm whether model context protocol access is included in the plan price.

Two passes across every retrieved surface located no hosting provider, no region, no country and no residency option. The security page discusses encryption and access control at length and never says where the data sits, which is the first question a buyer's own reviewer asks about a system holding mailbox contents.

The silence is not quite complete, and what fills it was not put there deliberately. Page metadata on the vendor's own site identifies a commercial application hosting platform serving the front end and a third party error monitoring service in use. Neither is disclosed anywhere in the security or privacy material, which means the subprocessor list a buyer can assemble from page source is longer than the one the vendor publishes, and the published one is empty.

For this product the stakes are set by what is ingested rather than by general principle. Mailbox and calendar contents, correspondence history with customers and investors, and notes about named individuals are among the more sensitive holdings a small vendor can accumulate, and a buyer in a regulated sector cannot complete a review without a processing location.

No single tenant option, no customer managed encryption key, no data processing agreement, no transfer mechanism and no published subprocessor list were located.

Ask in which country data is stored and processed, for the full subprocessor list including error monitoring and hosting, and whether any regional option exists.

A security and privacy page exists and describes practices in plain language, which is more than many vendors of this size manage. What it does not contain is a credential. Two dedicated passes located no service organisation control attestation, no international standard certification, no auditor name, no assessment period, no penetration test summary, no trust portal, no document request route and no subprocessor list.

The single audit related sentence states that continuous monitoring, vulnerability scanning and periodic audits are part of security practices. Under the credential test that is a noun with no verb and no scope boundary: who performs the audit, against which framework, at what frequency and with what outcome are all unstated, and an internal engineering review satisfies the sentence exactly as written.

One genuine credential does exist and is correctly formed, which is why this does not sit at the floor. The Google limited use undertaking names a specific policy, links to it, and commits to its requirements by name for data obtained through those interfaces. It is externally defined, externally enforceable and scoped to the vendor's most sensitive ingest path. A buyer should recognise it as covering the mailbox integration and nothing else.

The model context protocol page separately claims enterprise grade security and full audit logs that the security page neither describes nor supports, so the strongest security language on the site sits on the page least equipped to substantiate it.

Ask whether any external attestation exists or is in progress, and confirm the audit log claim.

One published price, one plan, no tiers, no seat minimum and no sales call: 29 dollars per user per month billed yearly, with the plan contents itemised as map view, filters and lists, network sync, import from a customer relationship system and priority support. A trial starts without a card. In a category where most competitors gate the number behind a demo, publishing a single figure is a real disclosure, and having only one plan removes the tier gaming that makes comparison difficult.

Three inconsistencies hold it below the top band, and they are inconsistencies rather than omissions, which is the more avoidable kind. The monthly rate of 39 dollars per user appears only in a question and answer entry on the home page and not on the pricing page, so the page a buyer visits to learn the price discloses the cheaper commitment and omits the more expensive option, which is the wrong way round. The trial length is given as seven days on the home page, fourteen days on the model context protocol page, and no duration at all on the pricing page. And the model context protocol server is presented as new, with its own get started route and its own trial language, without any statement of whether it is included in the single plan or costs extra, which is the first question a buyer interested in it will have.

Two passes located no annual commitment term, no refund policy, no cancellation provision and no statement of what happens to a workspace at the end of a trial.

Ask which trial length applies, whether the server is included, and what the cancellation terms are on the annual plan.

The strongest axis on this record, and it earns the top band on the framing as much as the substance. The security page states that at any time the customer can export all their information or delete their account, and that deleting the account or workspace permanently erases all records and associated data. Both halves of portability are addressed, retrieval and destruction, and both are described as actions the customer takes rather than requests the vendor may choose to grant.

That distinction is the whole point of this axis. Most records in this lane describe deletion on request through a support channel and say nothing whatever about getting data back out. Here the commitment is unconditional in time, unconditional in scope, stated on a page the vendor publishes rather than buried in terms, and phrased as a capability the user already has.

Two supporting elements reinforce it. Permissions on connected accounts are revocable at any time, so a customer can halt ingest independently of leaving, which is a partial exit most vendors do not offer. And the model context protocol server provides a second, programmatic route to the same records, so a technically capable customer has an extraction path that does not depend on an export button working.

The gaps are secondary and worth naming rather than grading down for: no export format is specified, no statement clarifies whether the export carries correspondence history and notes or only contact records, and no post termination retention period is given for backups.

Run an export during the trial and confirm what the file actually contains before committing the team's network to it.

Not applicable in the provider sense and rated accordingly rather than penalised. The platform transmits nothing. There is no sequencer, no bulk mail, no dialer and no messaging channel, so sender reputation, authentication records, domain warm up, throttling and bounce handling all sit in whatever tool the buyer uses to reach people. Grading this vendor down for absent sending controls would penalise it for a product boundary it drew deliberately.

What the platform does contribute is upstream of sending and genuinely relevant, which is why this sits mid band rather than being disregarded. Job change tracking keeps contact records current against the single largest cause of hard bounces in a stale database, which is people leaving the company the address belongs to. Consolidating every team touchpoint into one company level view prevents two colleagues contacting the same person in the same week, and the vendor names avoiding duplicated work explicitly as a benefit. Both are real inputs to a sending team's reputation even though the vendor never frames them that way.

What is missing is any quality claim behind them. Two passes located no email validation or verification capability, no bounce rate figure, no data freshness statistic, no statement of how quickly a job change is detected and no accuracy claim of any kind on the synced records.

Ask how quickly a job change is detected after it happens, and whether addresses are validated before they reach your sending tool.

The vendor is unusually willing to say what it is not, and that clarity is worth crediting. A published answer states plainly that pipeline software is organised around deals and stages while this product is organised around people, and that it has no pipeline, no deal stages and no lead scoring. Declining three capabilities buyers routinely ask for, in writing, on the marketing site, is a positioning choice most companies avoid, and it saves a mismatched buyer a wasted evaluation.

Intended users are named across sales, fundraising, partnerships and leadership, with a third party profile adding venture capital, consulting and executive teams and identifying small and mid sized teams as the target size. The absence of a seat minimum and of any enterprise tier is consistent with that.

The gaps are ordinary for a company at this stage: no named customers, no geographic coverage statement, no language support statement, no team size guidance and no vertical specialisation.

One internal tension is worth flagging because it cuts against the clearest positioning claim on the site. The assertion of no pipeline sits alongside a board view feature the vendor describes as letting teams create custom pipelines for deals, outreach or client engagement and move contacts between stages. Both statements are the vendor's own. A buyer who chose this product specifically because it is not pipeline software, or specifically because it does have a lightweight pipeline, should establish which description is accurate before committing.

Ask to see board view in a trial and judge whether it is a pipeline.