Glossary

Same word, different products#

Four words exist in every product and mean something different in each. Where these docs use them without a product name, they mean the Mindset version.

Terms#

Acceptance test

A situation you put an agent in, plus what must be true about how it behaved. Each test runs several times and passes on a threshold. By default a model-judged check runs five times, and a code check once. Also called a BAC (behavioral acceptance criterion). You write, run and retire them in Govern → Optimize. The agent's own Acceptance tests tab lists its tests, read-only. A test refuses every write the agent tries, so running one never changes your systems. See Test behavior.

Example: "Given an invoice with a $40 difference, the agent calls the invoice-comparison function."

Not: A gate the platform enforces. Activating a version doesn't run the tests. Once accepted, a test can't be edited. You amend it (which replaces it) or retire it with a reason.

Active version

The version of an agent that people get, in that environment. An agent has at most one. A new agent has none until you activate one, and until then it shows as Unavailable.

Agent

An agent has a prompt, a model, the resources it may reach, and optionally a script. People talk to it in the Hub, and schedules, triggers, other agents or your own systems can start it.

Example: The invoice exceptions agent compares invoices to purchase orders and registers supplier queries.

Not: A script. Most agents are instructions plus the right operations.

Agent Builder

The assistant docked on the left of an agent's tabs. Describe a change in words and it edits that agent. One of the builders.

Not: Orca. Orca provisions things you don't have yet.

Agree writes

The button on an agent's Chat tab that lets a saved version that isn't active perform its writes. Until an admin presses it, those writes are refused. The agreement covers a fixed list of writes.

API key

An org credential that lets your own systems call Mindset: start an agent through the API, or create sessions for the SDK. An org admin creates keys in Settings → API keys. Each key works in one environment only. See Environments.

Not: A model key. A model key is how Mindset calls a model provider. An API key is how your systems call Mindset.

$ask step

A script step that puts a decision to a person in Slack, with a button for each choice, and parks the run until they answer. It's the only way to get a person's yes on each individual action, because an operation's approval is given once. Slack is the only place it can post. See Make it follow steps.

Example: Finance approves each supplier query before the Prepare phase registers it.

Availability

Whether an agent can be reached, and from where. Set on the agent's Versions & Availability tab: one Available switch, plus the places it can be used (The Hub, Embedded in your own product, Other agents, Schedules, Triggers, and Outside systems). It applies as soon as you save and isn't versioned.

Builders

The assistants that create and edit one kind of resource from a description: the Agent Builder, Connection Builder, Function Builder and Widget Builder, plus the Audit Agent on the Audit page. Orca hands work to them. With Orca, they are the built-in agents, which on the shared deployment run on keys Mindset provides until an admin turns on Run the built-in agents on our own key.

Connection

Links Mindset to one outside system and holds the credential for it. The credential stays with Mindset and never reaches the agent or the model. The outside system sees the connection's shared credential. There are seven types: MCP server, Knowledge base, Postgres database, Google Sheet, HTTP API, StackOne, and AI model.

Example: The finance system, and the supplier contracts knowledge base.

Not: What you give an agent. You give it operations.

Embedding key

The OpenAI key Mindset uses to index and search knowledge bases, separate from model keys. Set in Settings → Embedding keys, with an optional endpoint to keep embedding traffic in a region you choose. Without one, the deployment's own key is used.

Environment

A named partition inside your org. Agents, connections, triggers and schedules each live in one environment. Every org starts with one called Sandbox. Org admins create more and can rename them. The slug in its address never changes, and an environment can't be deleted.

Not: A copy of your work. Nothing moves between environments on its own.

Experiment

A comparison run of an agent's acceptance tests with one change, such as the model or part of the prompt. It reports, per test, whether the change improved, regressed or made no detectable difference, and what each arm cost. It changes nothing on the live agent.

Function

A fixed list of steps that gives the same answer for the same input. An agent calls it as a tool. A step calls an operation, calls a model, or works on data it already has. Functions keep versions, must pass validation before they publish, and roll back with Roll back to this. Publishing or rolling back meets the go-live check.

Example: The invoice-comparison function fetches the PO, extracts both sets of lines, matches them and returns the differences and their total.

Not: Started by a schedule. Schedules and triggers start agents.

Gate

The condition a script phase must meet before the run moves on. A machine-checked gate tests what was collected, with no model involved. A judged gate asks a model to decide, and records its reasoning. A gate can combine both.

Example: Phase one passes when both the invoice and its purchase order have been retrieved.

Go live untested

The override an org admin uses when activation or publishing is refused because operations are untested. It's recorded with their name and an optional reason, and the operations stay Untested until a real call succeeds. See Versions, testing and going live.

Govern

The part of the Studio where you see what happened and improve it: Monitor (cost, performance, resources and topology), Log (every session, in order), Optimize (acceptance tests and experiments) and Audit.

Hub

Where people in your org talk to agents. An agent appears in the Hub for every member, in that environment, when it is available there: Available is on, The Hub is ticked and a version is active. There is no per-person list. Org admins can reach any agent, on the Hub or not. Members can also reach Hub agents from their own AI client over MCP.

Not: Where you build. That's the Agents area.

Lead agent and helper

A lead agent hands part of its work to another agent, its helper. The helper needs Other agents ticked in its availability, unless it's on the Hub and the lead agent is one a member is talking to. See Build an agent.

MCP

Model Context Protocol, a standard way for an AI client to see what another system offers and call it. Mindset uses it in two directions: it connects to remote MCP servers as a connection type, and it runs its own MCP servers so Claude and other AI clients can call your agents. See Use your agents from AI clients.

Model key

Your org's own credential for a model provider (Anthropic, OpenAI or AWS Bedrock). Set by an org admin in Settings → Model keys. On a self-hosted deployment, an instance key set by the operator sits beneath every org and serves any org with no key of its own. See Use your agents from AI clients.

Operation

A single action on a connection, with a fixed input and output, marked read or write. Operations are discovered from the system or written on the connection. An operation must be enabled before any agent can call it. With the org's auto-enable setting on (the default), a new write operation is enabled the moment it's added. See operation approval.

Example: "Get the purchase order for an invoice number" (read) and "register a supplier query" (write).

Operation approval

A write operation needs consent before anything can call it. Consent is given once per operation, not per call. The org setting Turn on connections that change other systems automatically is on by default, so new write operations are enabled without a person. When it's off, a person approves each new write operation on the connection's Operations tab. The buttons to stop one are Withdraw approval (a write) and Disable (a read). See Approve what agents can do.

Not: Up to an agent. No agent tool can approve or withdraw an operation. For a person's yes on each individual call, use an $ask step.

Orca

The agent that provisions what you don't have yet. Describe what you want, and it looks at what already exists, proposes a plan and stops for your approval. Specialist builders then create each piece.

Not: Where you edit something that exists. That's the Agent Builder.

Org

Your Mindset organization, and the isolation boundary. Members, settings, model keys and audit sit at org level. Its region comes from the deployment it lives on and can't change. See region.

Personal agent

An agent a member builds for themselves in the Hub. Off by default, and turned on in Settings → Defaults. A personal agent holds no connections or other resources. It can only hand work to agents already published to its owner, and it is never shared with anyone else.

Personal data

The org policy, in Settings → Governance, that replaces detected personal data with placeholder tokens before it reaches a model. Off by default. See Audit and personal data.

Phase

One step of a script, defined by what must be true before the run moves on. A phase says what the agent is told, what it Can use, what it Collects before advancing, and the gate to pass. Routes only go forward.

Example: Gather, Compare, Assess, Prepare.

Region

Where your org's data lives, set by the Mindset deployment it's on. It can't change after the org is created.

Rehearsal

The label on a run of an agent version that isn't active, started with Run now on the Triggering tab.

Run

One execution of an agent, from start to finish. Every run is recorded: what started it, the phases, the operations called, what came back, and any runs it handed to other agents. A run can park while a script waits, on a phase wait or on an $ask step in Slack, and resume when the answer comes.

Schedule

Starts an agent at set times, with instructions for that run. Set on the agent's Triggering tab with New schedule. Schedules are part of an agent's version, so a new or deleted schedule takes effect when that version is activated. The agent needs Schedules ticked in its availability. See Start it from your server or a schedule.

Script

The ordered phases an agent works through, for work that must wait or happen in order. A script belongs to one agent and has no history of its own: the agent's version records the script's text, so going back to an earlier agent version brings back the script as it was. See Make it follow steps.

Studio

Mindset's web interface, where people build, connect, run and govern agents. Builders and operators can do the same work from a coding agent or AI client through the MCP server. Its main areas are Agents, Orca, Connections, Functions, Widgets, the Govern pages (Monitor, Log, Optimize, Audit), the Hub and Settings.

System prompt

Tells the agent what it is and how to behave, on every turn. In the Studio it has five parts: Purpose, Personality, Background, Policy rules and Output formatting.

Not: The place for a rule that needs a definition. "A material discrepancy" belongs in a function as a threshold.

Test parameters

Stored inputs kept beside a script or a function, each with a Run button. Running one runs the script or function on those inputs and shows every step as exercised or not. A run that writes needs an admin's go-ahead for its write list. Successful calls make operations tested.

Tested and Untested

Whether an operation has had a successful call since it last lapsed, in this environment. It lapses when the operation's call shape changes, the connection's base URL or credential changes, or a call breaks its established output. Activating a version that puts new script text live, or publishing a function, is refused over untested operations. See Versions, testing and going live.

Topology

The tab under Govern → Monitor that shows which agents hand work to which, what each was granted and what it actually called.

Trigger

Lets another product start an agent by sending an event to an address Mindset gives you, signed with a secret. Set up as an inbound source and bound to the agent on its Triggering tab. The agent needs Triggers ticked in its availability. See Start it from your server or a schedule.

Unavailable

The badge on an agent that nobody but an admin can reach, because no version is active, Available is off, or no place is ticked. Hover over it to read which.

Version

A saved snapshot of an agent's configuration: the five prompt fields, model, reply language, name, script and its text, resources and schedules. Saving creates a version and activates nothing. Activating a version makes it live from each person's next turn, and rolling back is activating an earlier one. See Change something that's live.

Widget

A UI component an agent shows a person in a conversation instead of plain text. Widgets keep versions, must pass validation (it will render) before they publish, and roll back with Re-activate on an earlier version.

Write

An operation that changes something in an outside system. See operation approval.

Example: Registering a supplier query.

SDK terms#

These come up when you embed an agent in your own web app.

mindset-agent element

The drop-in web component, <mindset-agent>, that puts a complete agent chat panel on your page. It's the only way to show a scripted agent's greeting, cards, phases and questions. See The mindset-agent element.

Pass-through value

A value a tool needs that the model must never see, such as the signed-in user's account number. The page sets it, and the SDK adds it to the calls of tools that declare a parameter of that name. It lives in the browser, so it's never secret: never put a key, password or token in one. See Sessions and trust.

Session

The credential your backend gets from Mindset, with an API key, for one user and one agent, and hands to the browser unchanged. The SDK runs the agent with it. See Sessions and trust.

UI-less client

The SDK's JavaScript client with no interface of its own. It gives you the same agent and conversation as the element, with events in and commands out, for an interface you build yourself. See The UI-less client.