AI and connectors

Augos Factory data is structured at the source so that an AI assistant, an ERP or a BI tool can read it without clean-up. Every report exports as a signed AI Bridge payload or as CSV and Excel; the API returns the same measured data on demand; and a read-only MCP connector lets Claude, ChatGPT or Gemini query it with your own credentials.

Reviewed 2 October 2026 · Author: Tim Stevens, Augos · Source: Augos platform architecture and AI guidance

Structured at the source

Most plant data reaches an AI model after a long detour: exported from one system, cleaned in a spreadsheet, relabelled by hand, and then uploaded with a paragraph explaining what the columns mean. Augos removes the detour. The stop reason is captured by the operator at the Augos Operator Terminal at the moment it happens; every count is a machine pulse; every stage of a response is a system timestamp. The platform holds all of it in one vocabulary, the one on these pages, and every report is drawn from that single record. So the data a model reads is the data the report shows, and both trace back to the pulse they came from. AI reads the Augos record; the measurements themselves come from pulses and timestamps.

Source: Augos Factory method, How the platform works; Glossary: AI Bridge payload

The AI Bridge payload

The AI Bridge payload is the Augos name for a programmatic export built to be read by a machine: a language model, an ERP such as SAP or Sage, or a BI tool. It carries the report's figures and, alongside them, a short embedded guide to what each field means and how the fields relate, so the model does not have to guess at the shape of the data. Each Factory report payload is wrapped in a metadata envelope with a report identifier, a schema version, a SHA-256 checksum and an audit-trail reference back to the source pulses, and the schema is versioned per report type, so a payload produced today can be verified and re-derived later. A payload is scoped by the same point-level permissions as the platform: it contains only the measurement points the user who exported it is cleared to see.

Source: Augos Factory method, report envelope and signature; Glossary: AI Bridge payload; Roles and access

Exports from any report

Every report has an export menu. Choose the AI Bridge payload for a model or an integration, or CSV and Excel for your own analysis at the resolution the report is showing, down to the individual stop. A one-off question needs nothing more than this: export the month's Unplanned Stoppage analysis, hand the payload to a secure model, and ask which fault code cost the most on the bagging line and which stage of the response owned the minutes. The answer comes off your own measured events, and the source line on every figure lets you check it.

Source: Augos platform, report export menu; Performance Overview

The API

For a standing feed into your own systems the API returns the same measured data on demand, with no manual export. Access is granted per measurement point: email support@augos.io with the measurement points you need and Augos issues a token, sent as a Bearer token in the Authorization header of every request.

Measured data for a point and period:

Data returns as JSON, one object per interval, with dateUTC as the UTC start of the interval. The list of your points and their IDs comes from the points endpoint with the same header; set take or all=1 to receive rows.

Each point allows up to 48 data requests a day and up to 10 080 records per request; an additional API subscription lifts the limit. A token returns only the points it is cleared for, the same boundary a person has.

Source: Augos Data API reference; Roles and access

The MCP connector

The Model Context Protocol is the open standard AI assistants use to call external tools. The Augos MCP connector is a read-only connector for Claude, ChatGPT and Gemini, offered on limited access. Each user connects with their own Augos credentials, so the assistant sees exactly the measurement points, reports and periods that user is cleared for and nothing else. It can read; it cannot change a configuration, a standard or a record. Once connected, a plant manager can ask an assistant, in plain words, which line lost the most time to Inactive Wrench last week, how this month's Operator Discipline compares with last month's, or what the Best Operator gap is on a given line, and the assistant fetches the measured answer from the platform. Ask Augos about access.

Source: Augos platform architecture, MCP connector; access per Augos, October 2026

Using AI safely

Three rules from the Augos AI guidance. First, never upload raw Augos data to a public, consumer-tier AI service: consumer models may retain what you upload, and a minute-level production record reveals volumes, shift patterns and product mix. Use an enterprise model with a zero-retention agreement, a model hosted on your own infrastructure, or the MCP connector, which keeps the data inside the permission boundary. Second, strip sensitive metadata such as internal ERP codes or employee identifiers before passing a payload to a third party. Third, treat an assistant's output as analysis to be checked: every figure it quotes should be traceable to the report it came from, and with Augos data it is.

Source: Augos AI and security guidance; Data security and storage

Access

CSV, Excel and the AI Bridge payload are available from every report. API access is included in the standard fee. The MCP connector is on limited access; ask Augos about access. See What Augos Factory includes.

Source: Augos Factory single offer, 6 October 2026

Common questions