Error rendering macro 'rw-search'

null

Downloads

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 5 Next »

This page follows the data that iDNA Applications reads from your Notes environment: where it is stored, how it is processed, who sees it, and what leaves your network, AI clients included. It is for reviews that need more than What iDNA Applications Reads from Your Notes Environment, which describes the reading itself. It describes iDNA Applications 3.5.0 and later.

Overview

  1. Read: iDNA reads from your Domino servers with the collection ID (see What iDNA Applications Reads from Your Notes Environment).
  2. Stored on the appliance, a Linux host in your network.
  3. Processed every night on the appliance into the analyses.
  4. Shown to the people you give an iDNA login.
  5. Retrieved by AI clients, only if you enable the MCP server.

For reviews by a works council, a data-protection officer or a compliance function, panagenda provides two documents: the AI position paper (the iDNA Applications Companion and the MCP server), available from panagenda, and the note on user name pseudonymization, in English and German. This page summarises them; where a detail matters, the documents apply.

Storage on the Appliance

  • iDNA runs as a set of containers on a Linux host that you provide and operate in your network. Access to the host is yours to control.
  • Collected data is stored in a PostgreSQL database on that host, with real names. Pseudonymization is applied later, when the data is prepared for the analyses.
  • Files on the host include the design exports (compressed files) and the Notes runtime with the collection ID file and its password, which iDNA needs to collect without anyone logged in.
  • Snapshot data such as the directory, server documents and the catalog is replaced on each collection. Activity data such as session records is added over time.
  • Raw data that the nightly processing has already used can be deleted after a period that an administrator sets in the Portal under [Data Retention and Cleanup Options]. The processed figures are kept.
  • Backups, exports and copies of the host are yours to control. panagenda has no standing access to the appliance; access for support or consulting is granted by you.

Processing

The analyses are computed every night on the appliance; dashboards therefore show the data up to the previous night. The processing pseudonymizes user names as described in the note on user name pseudonymization. In short:

  • Pseudonymization is the default. User names in the per-user usage analyses (who used which application and when, per-user summaries, the license analysis and the editor statistics of the content analysis) are replaced with a pseudonym derived from the name. A person keeps the same pseudonym over time, so trends still work; the organisation part of the Notes name stays readable.
  • It is pseudonymization, not anonymization. iDNA has no function that turns a pseudonym back into a name, but anyone who holds the list of names and the method can rebuild the mapping.
  • Small departments and locations: while pseudonymization is on, a department or location with fewer assigned users than a threshold (15 unless the license sets another value) is shown under a placeholder. Its row and figures stay, so totals still add up.
  • Shown with real names on purpose: access-control entries, agent signers, group memberships, the people who last changed profile documents, who last accessed and who last wrote to each application, person documents from the Domino Directory, and MarvelClient data. These describe permissions, responsibilities and recent use; they are not intended for assessing anyone's work.
  • Switching it off takes a new license, which panagenda issues only after your organisation has confirmed in writing that it has taken the necessary internal steps, including with the works council. On a trial license the usage analysis is always pseudonymized.

Who Sees What

  • Inbound connections: a standard installation accepts browser connections on port 443 (HTTPS) and redirects port 80 to it. Behind port 443 are the Portal, the admin interface, the Metabase reports, DataMiner and, when enabled, the MCP server. Two more ports are optional and closed by default: VNC (TCP 5901), only while an administrator runs the VNC server on the appliance host for its graphical desktop, and PostgreSQL (TCP 5432), only if the database port is published for direct database access.
  • Logins: the Portal and the admin interface use local iDNA accounts or your LDAP directory, set up separately for each. Metabase has its own login. You decide who gets a login.
  • Roles: Portal users have a role (for example Admin, Editor or User) that decides which pages they can use; separate roles cover MCP and API access. The admin interface and the system pages are for administrators only.
  • Pseudonymization applies to the prepared data, so every view built on it shows the pseudonyms. Per-user views in the Portal show pseudonyms while pseudonymization is on.
  • Outgoing connections for logins and email: to your LDAP directory, if you use it for logins; to your mail server, for alerts and report subscriptions to the recipients you configure.
  • On a trial license some names and databases are masked or hidden; see iDNA Applications Trial License.

What Leaves Your Network

ConnectionFrom → toWhat is sentWhen
Software imagesAppliance → panagenda's download server (docker.panagenda.com)Requests for the iDNA container imagesOnline installation and upgrades. The offline installer needs no connection.
Update checkThe user's browser → panagenda's file server (files.panagenda.com)A request for a version file; no iDNA dataWhen the Portal login page is shown, unless the license has expired
Content analysis tokensThe user's browser → panagenda's licensing server (licensing.panagenda.com)The license key and the replica ID of the database to scan; the license key alone when the token status is shownWhen a user starts a scan of a replica set not yet authorized on the appliance, or opens the token status
Help linksThe user's browser → panagenda's documentationA normal page requestWhen a user clicks a help link
AI clientsAppliance → the user's AI assistant platformThe results of the tools the assistant callsOnly when the MCP server is enabled; see AI Clients and the MCP Server
Support bundlesAn administrator → panagenda supportThe appliance's configuration and log files; review the bundle before you send itOnly when an administrator downloads a bundle and sends it
  • The license is a file you upload; it is checked on the appliance. iDNA's own components send no usage statistics or crash reports to panagenda.
  • Metabase, the reporting component, has its own settings for anonymous usage tracking and for checking for updates. Check them in Metabase's admin settings.

AI Clients and the MCP Server

The MCP server lets an AI assistant that your organisation already uses retrieve analysis results from iDNA. The AI position paper describes it in full; in short:

  • Off by default. It is enabled through the license, at your organisation's request, and switched on by your administrator. Your organisation acknowledges the position paper when it opts in.
  • Restricted mode is the default: a fixed set of tools, no ad-hoc queries, no per-user analysis tool and no access to the tables that record individual activity. The AI sees application metadata, aggregates and source code. The names that Notes records as part of an application (for example access-control entries and agent signers) are readable by the server's database account, but no current restricted-mode tool returns them.
  • Permissive mode (ad-hoc queries and per-user analysis) must be chosen deliberately, as a separate license decision.
  • Access: every user gets a personal access token, valid for 365 days by default; every token has the same data scope. Every tool call is recorded in an audit log on the appliance.
  • Where results go: tool results travel over HTTPS to the user's assistant platform and into the AI model's context. From there, storage, retention and use are governed by your organisation's contract with the platform's vendor. In its default configuration the MCP server opens no connection to any external service.
  • Purpose: the AI components are intended for application portfolio analysis. They are not intended for monitoring or evaluating individual employees.

Setup: Setting up the MCP Server and Setting up iDNA Applications Companion.

Related Topics