HRDS001A – OCTOBER 2026 – Rev A · Oct 2026

HRD-1 § Datasheet Book a call

PreviewAgentsInferenceCompany brain

HRD-1 Applied AI module with agents, inference and an integrated company brain

Herendil helps companies put AI to work on their own problems. We advise on where agents and models fit, then build them into your tools, your data and your team.

Figure 1. HRD-1, shown in a QFN package for illustration. Ships as people and code.

1 Features

  • Applied AI consulting and software, in one engagement
  • Agents that act inside your tools
    • Run under your permissions
    • Log every step for review
  • Inference matched to your cost, latency and privacy limits
    • Hosted, self-run, or routed between both
  • Company brain: shared memory for people and agents
  • Drop-in: works with your team and your stack
  • Engagements sized in weeks
  • Low overhead: few meetings, fewer slides

2 Applications

  • Support and operations teams answering the same questions
  • Knowledge spread across docs, chat and tickets
  • Agent pilots that need to do real work
  • Choosing between hosted and self-run models
  • Deciding where AI should not be used

3 Description

HRD-1 is how Herendil works with a company. It starts with the work: which questions, decisions and chores an agent or a model could take on, and which should stay with people. Then it builds that, inside the tools you already run.

Three blocks do most of the work. Agents act in your systems under your permissions and log each step. Inference is set up for your limits on cost, latency and privacy. The company brain makes what your company knows, decides and does available to people and agents alike.

HRD-1 is pin-compatible with existing teams. There is no platform to adopt.

Device information(1)
Part numberPackageBody size
HRD-1People and codeSmall team

(1) For all available configurations, see Ordering information.

4 Functional block diagram

Block diagram. Your data, people and tools feed the company brain inside HRD-1. The brain connects to agents and inference. Agents produce actions in your tools; inference produces answers for people and agents. DATAdocs, tickets, DBs PEOPLEyour team TOOLSyour stack 123 DATAPPLTOOLS COMPANY BRAINwhat you know, decide and do AGENTSact in your tools INFERENCEruns the models ACTANSHRD-1 45 6GNDcommonsense 7NCreservedfor hype ACTIONSin your tools ANSWERSfor people, agents
Figure 2. Dashed lines carry signals. Everything inside the box runs in your environment.
4.1Pin functions
PinNameI/ODescription
1DATAInDocuments, tickets, databases and chat. Read where they live; copied only when needed.
2PPLIn/outYour team. Sets goals, reviews output and owns the decisions.
3TOOLSIn/outThe systems you already run, with their existing permissions. HRD-1 connects to them and replaces none.
4ACTOutAgent actions inside your tools. Each one logged, each one reversible where the tool allows.
5ANSOutAnswers grounded in the company brain, with sources, for people and for other agents.
6GND—Common sense. Tie to every decision.
7NC—No connection. Reserved for hype; leave floating.

5 Specifications

TA = your office, VDD = your existing stack, unless otherwise noted. Typical values describe how an engagement is meant to run. They are not measured results.

5.1Typical characteristics, by engagement mode
ModeWhat it doesDurationYou get
ASSESSMap where AI should do work in your company, and where it should not.WeeksA written plan, ranked by value and effort
BUILDShip one agent that does one job inside your tools, with a person in the loop.WeeksWorking software in your repository
INFERChoose, host or route models to fit your cost, latency and privacy limits.WeeksAn inference setup your team can run
BRAINConnect what the company knows so people and agents can find and use it, with permissions intact.Weeks to monthsA company brain your agents can query
ADVISEStanding counsel for the team doing the work.OngoingA regular call and written answers
5.2Recommended operating conditions
ParameterValue
Engagement lengthweeks, not quarters
Interfaceyour team, your stack
Herendil team sizesmall
Code and data ownershipyours
Locationremote; on site when useful
Output formatsoftware, written decisions
5.3Absolute maximum ratings(1)
ParameterMax
Hype words on this page0
Slides before a working prototypeas few as possible
Vendor lock-innone
Data leaving your boundarywhat you approve

(1) Stresses beyond these ratings may cause permanent damage to trust. Operation at these limits is not recommended either.

6 Application notes

Drafts in progress. Titles may change before publication.

7 Ordering information

To order, book a call. We talk about your problem and whether HRD-1 fits it. No procurement portal, no sales sequence.

Book a call

The booking link goes here soon. Herendil is taking a small number of calls.

HRD-1-BUILD
│   │ └── Mode: ASSESS, BUILD, INFER,
│   │           BRAIN or ADVISE
│   └──── Revision
└──────── Herendil

Figure 3. Device nomenclature.

7.1Orderable part numbers
Orderable partStatusConfiguration
HRD-1-ASSESSPREVIEWAssessment and written plan
HRD-1-BUILDPREVIEWOne agent, built and handed over
HRD-1-INFERPREVIEWInference setup
HRD-1-BRAINPREVIEWCompany brain
HRD-1-ADVISEPREVIEWOngoing advice

8 Revision history

RevDateChanges
AOct 2026Initial release. Device status: preview.
—BeforeCompany dormant. Nothing to report.