Quillreach

One model capability is marketed: drafting connection notes and follow ups against the target's real headline, role and recent activity, in a voice matched to the account holder. Remove it and the product still stands in full, because everything that makes this vendor distinctive is scheduling and enforcement rather than generation.

Per action caps, the cap checking scheduler, warm up pacing, working hour windows, the hourly safety monitor, automatic pausing, branching sequences, stop on reply, multi account management and the unified inbox all survive untouched. What is lost is the drafting, and the customer would write the messages themselves as they did before. Modest claim, proportionate to what was built.

The strongest single oversight statement graded in this index, and it works by removing the customer's own authority rather than granting it. Each connected account carries a separate daily cap for each action type, the scheduler reads cap status before queuing every send, and when a cap would be exceeded the send waits for the next window. The vendor states plainly that this cannot be overridden from a setting and that it will refuse to send even when the customer asks it to.

A guardrail the buyer cannot switch off is a materially different thing from a default they can. Beside it sit four more mechanisms, all published: model drafts enter an operator review queue and wait for approval rather than sending, sends occur only inside working windows the customer defines, an hourly monitor watches acceptance rates, volume spikes and account checkpoints, and an account pauses itself on an anomaly and then states the reason. Sequences stop the instant a person replies. Off the top only marginally: no audit record, role model or permission set is described anywhere.

The drafting capability is marketed on three separate pages and no provider, model, family, version or hosting arrangement is named anywhere. One aspect deserves specific mention because it goes unaddressed rather than merely unnamed: the product writes in the account holder's own voice, which implies the system is conditioned on that person's prior writing, and nothing states what is used to establish the voice, where it is held, or whether it persists after the seat is cancelled.

This sits at the floor of the band rather than below it, and the distinction matters. There is no named customer, no logo, no testimonial, no case study, no customer count and no quantified result anywhere on the pages read, and the people shown inside the product illustrations are clearly invented personas with generic image filenames.

What keeps it out of the bottom band is that nothing is fabricated to fill the gap and the vendor declines the claim explicitly: asked what reply rates to expect, the safety page answers that reply rates depend on quality, which is a refusal to publish a number rather than an invented one. A vendor with no proof that also invents none is in a different position from a vendor whose proof is manufactured.

Genuine restraint is published and none of it is expressed as compliance. Sequences stop on reply so nobody is messaged after answering, sending is confined to working hours the customer sets and never runs overnight, and per action caps are enforced below the platform's own thresholds. Those are behaviours a compliance position would produce, arrived at for account safety reasons instead.

What is absent is the position itself: no statute, regime or regulator is named on any page read, no consent standard is stated for approaching people who did not ask, no suppression list is described, and no route exists for a recipient who wants no further contact other than replying, which stops one sequence rather than adding them to anything. Terms of service and privacy policy exist at their own routes and neither was opened.

A privacy policy, terms of service and refund policy each exist at their own route and none was opened, so this row is graded on the surrounding surface and flagged for re verification. One control is stated clearly and is worth crediting: account connection runs through a session based flow and the vendor states it never stores the password.

What the platform does hold is substantial and unaddressed on the pages read, being an authenticated session against a named individual's professional network account, the full message history of every conversation conducted through it, and enriched records on every imported lead. No retention period, processing addendum, subject request route or sub processor list appears anywhere.

Leads arrive from the customer's own comma separated file or from a search they ran themselves on the professional network, so the sourcing baseline is the customer's, which is clean. One step sits outside it and is unsourced: the platform states that it cleans, deduplicates and enriches every imported lead before anything is sent, and no enrichment source, partner, field set, coverage or accuracy claim appears on any page read. Enrichment that adds fields to a person's record from an unnamed source is the provenance question this axis exists for, and it is unanswered.

The most interesting tension in this build, and it resolves in the middle rather than at either end. Everything on the sending side runs the opposite way to this category's norm: one connected account per seat rather than rotation across many, caps set below the platform's own limits and unoverridable, warm up pacing, working hours only, and automatic pausing when the account shows a checkpoint.

The vendor is selling staying under the line, not crossing it, which is the reverse of the posture that earns the bottom band. And the architecture underneath is nonetheless built for non detection, stated openly. Browser extensions are rejected specifically because the platform can fingerprint their signature, actions run server side at randomised intervals rather than from the customer's own session, and free proxies across fifty or more countries sit on the price card as an included feature.

A proxy in the target geography exists to make automated activity look locally originated. Restraint in volume and evasion in transport are different things, and this vendor does both. No conformance position for the platform being automated appears anywhere.

Nothing published addresses what happens to the material the model works with. Drafts are generated from a target person's headline, role and recent activity and from the account holder's own writing style, and no statement covers whether either is retained, whether it trains anything, whether content crosses between customer workspaces, or where inference runs. The security page exists at its own route and was not opened, and is flagged as the likely home of any answer. Set against a platform holding live authenticated sessions for named individuals' professional accounts, the absence of any stewardship statement on the marketing surface is the gap.

Messages composed by a model arrive from a named individual's real professional account, written in that individual's voice, referencing the recipient's recent activity, with nothing indicating machine authorship and no position on the European marking obligation. Voice notes appear in the sequence step list and whether they are recorded or generated is not stated, which leaves the sharpest version of the question open. Two things hold this at the middle rather than lower.

Every draft passes through an operator review queue before it sends, so a person reads and approves the words that go out under their name, which is a materially different situation from unattended generation. And a reply stops the sequence immediately, so the recipient is not repeatedly approached by a system after signalling a human is present.

The thinnest connectivity surface of any vendor graded this session. Configurable webhooks per workspace and comma separated import are the whole of it. No system of record integration is named, no automation platform, no messaging platform, no marketplace, no documented interface, no developer reference and no protocol server.

For an outreach product that is a real structural gap rather than a missing convenience: every reply the platform earns lands in its own unified inbox and there is no published route for those conversations to reach the pipeline the customer actually forecasts from. The free proxy network across fifty or more countries is infrastructure rather than integration.

Hosting provider, country, region, data centre and residency option are absent from every page read, and no corporate address, entity name or jurisdiction is published anywhere. The company is not anonymous, since the founder is named and the footer links his personal site, but its location is not stated.

One published figure touches geography and answers a different question: proxies are offered in fifty or more countries, which describes where traffic appears to originate rather than where data rests. The security page was not opened and is flagged.

A dedicated security page exists in the footer and was not opened, so this is flagged for re verification rather than settled. What the rest of the site establishes is one real control and one architectural claim. The vendor states that account connection runs through a secure flow and that it never stores the customer's password, which for a product holding professional network sessions is the single most relevant thing it could say.

Actions run server side through dedicated session based infrastructure rather than the customer's browser. Past those, no certification, audit report, attestation, penetration test, encryption statement, trust centre, status page or vulnerability disclosure route was located on any page read.

One plan with every feature in it, priced at fifty nine dollars per seat per month for one to nineteen seats and thirty nine dollars per seat from twenty, with annual billing stated as two months free against monthly. The trial is fourteen days with no card required and cancellation is at any time, with a refund policy published as a document of its own rather than described in a sentence.

Nine included capabilities are enumerated on the price card, including the ones competitors commonly gate, and the vendor states that safety is the default rather than a plan upgrade. What earns the top band is a commitment most vendors in this index would not make in writing: the number on the pricing page is the number on the invoice, with no surprise renewals and no request to contact sales for the listed price. A buyer at any size can compute their exact bill and knows there is no ladder above them.

Configurable webhooks per workspace are the only published route by which anything leaves the platform, and they carry events forward rather than exporting a record backward. No export function, file format, post termination right, deletion timeline or retention period appears on any page read.

What accumulates and has no stated way out is the conversation archive: every thread from every connected account, held in the platform's own inbox with no system of record integration published to mirror it anywhere else. A refund policy is published, which addresses money rather than data. Terms of service unopened and flagged.

The channel is a professional network rather than email, and the discipline transfers directly. Published: separate daily caps per action type covering connection requests, messages, direct mail, reactions and profile visits, a scheduler that checks cap status before every queue, gradual ramping for new accounts, sends restricted to customer defined working hours, randomised intervals rather than fixed timing, hourly monitoring of acceptance rates and volume spikes, automatic pausing on anomaly with a stated reason, and sequences that stop on reply.

The vendor also explains the underlying asymmetry rather than asserting safety, noting that a flagged account loses six to twelve months of accumulated standing and a banned one is unrecoverable, so one day of over sending costs the whole account. Holding it off the top band, not one cap figure is published anywhere, so a buyer cannot compare the enforced ceiling against the platform's own limits, and no acceptance rate floor or degradation response threshold is stated.

Three buyer types each have a dedicated page, covering founders, agencies and sales development teams, and the seat structure implies the intended range with a volume rate beginning at twenty seats and a one account per seat rule that fixes what a seat means. A comparison page against named alternatives exists.

Against that, no customer is named in any sector, no customer count is published, no region, language or industry coverage is stated, and there is no statement of who the product is wrong for at a vendor whose whole argument is that some buyers want volume it will refuse to provide. That refusal is the obvious basis for a boundary statement and none is written.