TRANXACT.AI Help
Agentic AI › Build

Agents

What each agent is for, what it may do, and what it has actually been doing. An agent with no tools is not broken — it is one that can only think, which is a legitimate and common thing to want.

Where to find it

  1. Module: Agentic AI
  2. Group: Build & Orchestrate
  3. Section: Build
  4. Screen: Agents

Address in the console: /agentic/agents

List · 23 columns · 25 fields on its form · 5 of them required · 5 actions

What the list shows

These are the columns of the list, in the order they are drawn. Any of them can be sorted on, filtered on, hidden, or exported.

ColumnKindField name
AgentCodeagent_no
NameTextname
ModuleStatusmodule
KindStatuscategory
AutonomyStatusautonomy
StatusStatusstatus
On or offStatusswitched
OwnerTextowner_name
ModelCodemodel
EngineTextengine_label
Model problemTextmodel_note
Model used (last run)Codemodel_used
Changed since publishYes or nounpublished_changes
ToolsNumbertools_granted
Can changeNumberwrite_tools
Runs (30d)Numberruns_30d
SuccessNumbersuccess_rate
WaitingNumberawaiting_approval
Cost (30d)Amountcost_30d
Records changed (30d)Numberrecords_30d
Produced (30d)Numberproduced_30d
IncidentsNumberopen_incidents
Last runDate and timelast_run_at

The form

What is asked for when a record is created here. Required fields are marked; a choice drawn from a configuration list is one an administrator can extend without a release.

FieldTypeChoicesNotes
Code required Text Short and stable. It appears in the audit trail against everything this agent does, so it cannot be changed once the agent is created.
Name required Text
Name in local language Text The agent's name in the company's local language — Arabic in Saudi Arabia, Hindi or a regional language in India. People using the console in a language other than English see this name; English readers see Name. Leave empty to show Name to everyone.
What it is for Long text One or two sentences a colleague reads to tell whether this is the agent they need.
Module required Choice agentic, crm, scm, accounting, hrms, mfg, ats, ehr, slcm, platform The part of the product it works in. Its tools are usually from the same part.
Kind of work required Choice Configuration list What sort of work it does, for grouping and filtering agents. Starts at Operations; choose the one that fits.
Instructions required Long text What it is, what it must do, and what it must never do. This is the whole agent: everything else on this screen is a limit on it.
Opening instruction Long text What it is told at the start of every run, as distinct from what it always is.
Autonomy Choice Configuration list Suggest only is where every agent starts. Raising this is a separate, audited decision — the field here sets it on a new agent, and afterwards it changes on the agent screen.
Runs as Choice Looked up from existing records The agent borrows this person authority and cannot exceed it. An agent that runs as an administrator can reach everything an administrator can.
Owner Choice Looked up from existing records The person answerable for what this agent does. Left empty on a new agent, it is you. Needed before the agent can be published or given more autonomy.
Engine Choice Looked up from existing records The AI service it runs on. Starts at the company default engine (Platform › AI Providers); an engine with no key for this company cannot be chosen.
Model Choice Looked up from existing records Required before the agent can be published or run. The list is the models of the chosen engine; a model with no price on file shows its run cost as unknown.
Temperature Number How freely it words its answers: 0 is the most predictable, higher is more varied. Empty: the engine default.
Reply size limit Number The longest reply it may write in one step, in tokens (a token is about three quarters of a word). Empty: the engine default.
Step limit Number A run that hits this is stopped rather than truncated, because a half-finished agentic run that claims to have succeeded is worse than one that says it ran out.
Tool-call limit Number The most tool calls one run may make; the run is stopped at this many.
Time limit (seconds) Number How long one run may take before it is stopped, e.g. 300 for five minutes.
Cost ceiling per run (company currency) Number
Cost ceiling per day (company currency) Number
Memory Choice Configuration list Whether it keeps notes between runs (Agent Memory) and how widely they are shared.
Forget after (days) Number A note it keeps is forgotten after this many days. Empty: kept until removed.
Writes in Text A language code. Empty means the language of whoever is reading.
Notes Long text
Enabled Yes or no Untick to switch the agent off without retiring it — it then does not run, by hand or on a trigger. This does not make a draft runnable: Publish does (Next step › Publish), and Status then reads active. Next step › Switch off / Switch on does the same without opening Edit.

What you can do here

Run it primary

On each row of the list. Offered only on records where it applies.

Run this agent now

Asks for: Anything it needs

Publish

On each row of the list. Offered only on records where it applies.

Saves the instructions, engine, model, autonomy and tool grants as a numbered version. Runs use the last published version until you publish again.

Asks for: What changed*

Switch off

On each row of the list. Offered only on records where it applies.

Switch this agent off

Asks for: Reason

Switch on

On each row of the list. Offered only on records where it applies.

Switch this agent on

Asks for: Reason

Change autonomy

On each row of the list.

Change how far this agent may act on its own

Asks for: Autonomy* Reason

* required

The rules that apply here

Everything on this screen obeys the platform rules rather than rules of its own. The ones worth knowing before you use it:

Beside it in Build