Engine Documentation
How the Agent Pricing Engine works: the concepts, the roles, what costs a credit, and how rule changes are governed. Self-serve sign-up is launching soon — until then, watch the demos or book a walkthrough.
Core concepts
Workspace
Your company's private instance, created from your work email domain. Colleagues who sign up with the same domain join it automatically. Each workspace has its own engine database — rule packs, quotes, and audit history are isolated per company.
Rule pack (product model)
The variables, allowed values, constraints, and pricing logic for one product family. The pack is the source of truth: every question the engine asks, every option it blocks, and every price it returns traces to a rule in the pack. Packs are versioned; quotes record the pack version they were priced against.
Configuration session
A conversation toward a quote. The engine ranks which question to ask next by how much it narrows the result. Answers can be firm or assumptions ("not sure, assume 480 V") — assumptions are flagged on the finished quote for confirmation.
Deterministic pricing
Identical facts produce identical quotes — always. The engine never improvises a number. If a combination is invalid, it says why and lists every minimal way out.
What's free vs. what costs a credit
| Action | Cost |
|---|---|
| Configuration sessions, questions, answers | Free, unlimited |
| What-if exploration, option enumeration, optimization | Free |
| "Why not X?" explanations and conflict diagnosis | Free |
| Rule proposals, impact previews, approvals | Free — governance is included |
| Submitting a quote (incl. routing, review packet, audit trail) | 1 credit |
| Re-working a rejected quote in the same session | No additional credit |
See pricing for tier allowances.
Roles
- Rep — runs sessions, submits quotes. Approval routing is decided by policy, not by the rep.
- Reviewer / manager — receives a review packet per submitted quote: the facts, decisions, price build-up, flagged assumptions, and any disagreements, with history.
- Engineer — teaches the engine in plain language. A proposed rule is previewed against the workspace's quote history (see exactly which past quotes it would have changed) and requires approval by a second engineer before going live.
- Admin — manages people, roles, approval policies (e.g. who approves discounts beyond the band, large orders, or quotes relying on unverified rules), and workspace controls.
Connecting AI assistants (MCP)
The engine is a full MCP server. Connect it to Claude or any MCP-capable assistant and your team works conversationally — "quote three freeze-protected units for a Class I Div 2 site" — while every configuration check and price still comes from your rule pack. The assistant talks; the engine decides.
Connection instructions (OAuth) ship with self-serve launch.
Integrations
Quotes sync to your CRM and post toward your ERP — Salesforce, NetSuite, SAP, Dynamics 365, HubSpot — via Nteli's integration practice. See ERP integration and CPQ & Revenue Cloud.
Importing your catalog
Rule packs can be authored in-app or imported from your existing price books; your historical quotes can be replayed against the pack to validate it before go-live. Enterprise plans include onboarding help — pricing.