Quick Verdict
Prox is not a generic support bot. It turns manuals, specifications, installation guides, fitment tables, and expert knowledge into a cited product system for complex manufacturers. Customers can configure products, field technicians can troubleshoot, dealers can answer questions, and internal teams can query the same knowledge through product widgets, team tools, and MCP.
Procurement must resolve a material public pricing conflict. At the July 21, 2026 cutoff, the marketing page listed a $2,000/month Launch plan, while the legal terms referenced a $5,000/month base platform fee. An order form may override either statement. This page therefore classifies Prox as custom paid rather than presenting one number as dependable.
Best For
- Manufacturers with complex catalogs, specifications, service procedures, and dealer networks.
- Support and training teams preserving senior technician knowledge.
- Organizations able to provide subject-matter review, role-based permissions, and change approval.
- Not ideal for generic chat, poor source documentation, or small teams unable to support implementation and governance.
Key Features
- Page-cited answers: installation, service, and troubleshooting responses link back to manual pages.
- Configure and quote: selectors, calculators, sizing tools, and kit builders support presales work.
- Visual answers: generated troubleshooting views, step guides, product cards, and diagrams.
- Multiple surfaces: product pages, dealer devices, Slack or Teams, service desks, and AI tools.
- Manufacturing knowledge capture: indexes internal questions and experienced staff knowledge.
- Remote MCP: reads catalogs and knowledge trees and exposes
edit_wiki_nodefor versioned text replacement.
Use Cases
A buyer can check compatibility on a product page, a field technician can retrieve cited service steps, and sales can query a catalog from an AI assistant. Knowledge administrators can correct nodes through the product or MCP. Treat MCP write access as a separate high-risk capability: default to read-only, approve writers by role, and require two-person review for safety specifications, service torque, medical devices, firearms, fire protection, aerospace, and other consequential content.
Pricing
Public materials describe a 30-day trial and usage allowances, but the homepage and legal terms conflict on the base monthly fee. Obtain an executed quote covering indexed pages, message definitions, overages, products, all deployment surfaces, MCP, SSO, support, accuracy reviews, implementation, renewal increases, taxes, and export. A marketing calculator is not a substitute for an order form.
Pros
- Manufacturing-specific workflows go beyond generic retrieval chat.
- Manual-page citations make expert verification more practical.
- Customer, dealer, internal, and MCP surfaces can share one knowledge source.
- The vendor claims SOC 2 Type II controls, tenant isolation, SSO, roles, and audit logs.
- Terms say customers own Customer Data and that it is not used to train general-purpose foundation models.
Cons
- The
$2,000marketing price and$5,000legal reference conflict. - SOC 2, encryption, and no-training positions are vendor claims that still require report, DPA, subprocessor, and exception review.
- A cited answer can still be wrong when a manual is stale or parsing fails.
- The MCP connector includes a write tool, increasing accidental-change and prompt-injection impact.
- Results depend on source quality, permission mapping, expert review, and continuous maintenance.
Alternatives
| Tool | Best for | Main advantage | Main limitation |
|---|---|---|---|
| Glean | Enterprise-wide search | Broad connectors and permission-aware search | Not focused on manufacturing configuration and repair |
| Intercom AI | Digital customer support | Mature channels, inbox, and automation | Industrial manuals are not the core specialization |
| Zendesk AI | Existing Zendesk organizations | Tight service-desk workflow | Manufacturing knowledge requires additional design |
| ChatGPT | General analysis | Flexible and broadly adopted | Requires custom retrieval, permissions, citations, and governance |
FAQ
What is Prox’s public price?
There is no single dependable public figure. Marketing and legal pages show different base amounts, so use an executed order form.
Does Prox train on customer manuals?
The site and terms say Customer Data is not used to train general-purpose foundation models. Review the DPA, subprocessors, de-identified data, and contractual exceptions.
Does SOC 2 Type II prove complete security?
No. It evaluates scoped controls over a period. Buyers should inspect the report scope, exceptions, and complementary controls.
Is the Prox MCP read-only?
No. edit_wiki_node can write a versioned correction, so write permission and approval need separate governance.
Can it autonomously guide safety-critical repair?
It should not replace qualified review. Verify the original manual, revision, model, and field conditions, with mandatory escalation for consequential decisions.
Bottom Line
Prox has a clear vertical thesis: cited manufacturing knowledge delivered wherever customers and staff work. That usefulness also raises the stakes because answers can shape configuration, quotes, and repair, while MCP can write knowledge back. Resolve price and data terms first, then test real documents, dangerous errors, permissions, approval, and rollback before production.