# Glossary — the words the Agents guide uses

> Every term the guide uses, in a sentence or two of plain words, each with a link to the page that explains it in full: agent, skill, tool, account, sponsor, piece of work, receipt, record, proposal, request, inbox, queue, standing rule, budget, meter, delegation, process, run, assignment, criterion, work repository, correction, instruction, inspector, policy, kit and bridge.

<!-- id: agent-guide-glossary · area: agent · stability: preview · html: https://osysharp.com/reference/agent/guide-glossary/ -->

> **You are here:** [Building with agents](https://osysharp.com/reference/agent/guide/) › Glossary

## Summary        {#summary}
The words are in the order you meet them, not in alphabetical order: the agent first, then its work, then the
people around it, then what holds work together over days.

## Description    {#description}

### What is an agent?   {#agent}
A part of an app that uses an AI model to do a job. In Osy# it is a class: a purpose, the jobs it can do, and what
it may spend. It has an account of its own in the app. See [Writing an agent — a class with a purpose, skills and an account of its own (Agents kit)](https://osysharp.com/reference/agent/agents/).

### What is a model?   {#model}
The AI service the agent thinks with, such as Claude, GPT or Gemini. The app lists the models it has; every call to
one is limited and counted. See [The Llm kit — call a model, and know what each call cost](https://osysharp.com/reference/agent/llm-calls/).

### What is a skill?   {#skill}
One job an agent can be given: what goes in, what comes back, and which tools it may use to get there. The result
is checked before the app sees it. See [A skill — a typed job an agent can be given, its check and its tools (Agents kit)](https://osysharp.com/reference/agent/skills/).

### What is a tool?   {#tool}
One of the app's own functions, offered to the agent. "Look up the order", "set the fee". The compiler works out
whether a tool only reads, changes data, sends something or spends, and the agent's permissions follow from that.
See [[Action] — what an app offers to be done, by name](https://osysharp.com/reference/function/action/).

### What is the agent's account?   {#account}
The identity the agent acts under. It holds exactly the roles its class names. Everything the agent changes is
written down as changed by that account. See [[Actor] — a principal that code may run as](https://osysharp.com/reference/security/declared-actor/).

### Who is "the person the work is for"?   {#for}
Whoever asked the agent, or whoever the work was given on behalf of. The agent can never read or do more than that
person could themselves. In a process nobody may be signed in; then the work is for nobody and the agent has only
its own rights. See [[agent-ask#for]].

### What is a sponsor?   {#sponsor}
The person who answers for an agent. An agent without one does not run. Questions and proposals that have nobody
else to go to go to the sponsor. See [[agent-agents#who]].

### What is a piece of work?   {#work}
One job given to one agent, from the moment it is given until it ends: done, failed, stopped or called off. It may
wait for a person in the middle, for days. See [Giving an agent work — Do, Ask, Stream, For and Decide (Agents kit)](https://osysharp.com/reference/agent/ask/).

### What is a receipt?   {#receipt}
The one-line summary every piece of work leaves: who asked, as which account it ran, what it was about, what it
cost against its limit, and how it ended. See [[agent-task-log#receipt]].

### What is the record?   {#record}
Every step of a piece of work, in order, each as a plain sentence: each model call, each tool, each wait, who
decided what. The lines are chained together, so a line changed afterwards is found. See [[agent-task-log#record]].

### What is the gate?   {#gate}
The check that stands before every tool an agent wants to use. It answers one of three things: do it, propose it
to a person, or never. See [[agent-autonomy#gate]].

### What is a proposal?   {#proposal}
Something an agent wants to do and may not do alone, kept exactly: what, to what, why, and what the agent had
looked at first. A person approves or rejects it. Approving runs it with that person's rights. See
[[agent-autonomy#proposal]].

### What is a standing rule?   {#standing-rule}
A person's "from now on, do this alone", for one tool of one agent, in the work it does for them. It can be taken
back, and it counts how often it was used. See [[agent-autonomy#standing]].

### What is a request?   {#request}
Something put to a person: a question with its possible answers, an approval of one exact thing, or a form. The
answer comes back to whatever was waiting for it. See [Requests to people — a question, an approval, a form, and the answer coming back (HumanRequests kit)](https://osysharp.com/reference/agent/human-requests/).

### What are the inbox and a queue?   {#inbox}
The inbox is the screen where a person finds what is waiting for them. A queue is a named pile of such items that a
group of people works: "Finance", "Purchasing". See [The WorkQueues kit — queues, who holds what, and the service levels promised on any item](https://osysharp.com/reference/workflow/inbox-kit/).

### What is a budget?   {#budget}
A hard limit in money: for one piece of work, for an agent in a day, for each run of a process, for a person, for
the whole app. A call that would go past it is not sent. See [What an agent may spend — a hard limit on every piece of work (Agents kit)](https://osysharp.com/reference/agent/budgets/) and [Model spend — budgets, the meter and the price book (Llm kit)](https://osysharp.com/reference/agent/llm-spend/).

### What is the meter?   {#meter}
The count of every call to a model: which model, how much text went in and out, what it cost. It keeps no text. See
[[agent-llm-spend#meter]].

### What is delegation?   {#delegation}
An agent handing a typed piece of its job to another agent, called a specialist. The agent in charge is the
coordinator. A specialist never gets more rights or more money than the work above it has. See
[An agent hands work on to other agents — at once, typed, and never with more than it holds (Agents kit)](https://osysharp.com/reference/agent/delegation/).

### What is a process, and what is a run?   {#process}
A process is a piece of business with fixed steps: qualifying a supplier, handling a return. In Osy# it is a
[workflow](https://osysharp.com/reference/workflow/index/). A run is one time through it, for one supplier or one return. See
[Processes — every run of a business process, listed and read from start to now (Processes kit)](https://osysharp.com/reference/workflow/processes-kit/).

### What is an assignment?   {#assignment}
A piece of work somebody asked for, owned until its outcome is reached. It says what "done" means as a list of
criteria. Agent work, requests and processes are attached to it and may fail or be replaced; the assignment stays.
See [Assignments — a piece of work owned until its outcome is reached (Assignments kit)](https://osysharp.com/reference/agent/assignments/).

### What is a criterion?   {#criterion}
One thing that must be true for an assignment to be done, with what showed it was met. See
[[agent-assignments#done]].

### What is a work repository?   {#repository}
What an assignment knows, kept as items: evidence, letters, findings, decisions. An item is never overwritten; a
change is a new version that says who wrote it and from what. See [Work repositories — what an assignment knows, kept as items (WorkRepositories kit)](https://osysharp.com/reference/agent/work-repositories/).

### What is a correction?   {#correction}
A person saying, on one piece of work, what was wrong, what right looks like, and why. It changes nothing by
itself. See [[agent-teaching#correction]].

### What is an instruction?   {#instruction}
Something an agent is told from now on, drawn from corrections. It is only proposed until a person who teaches the
agents makes it active. Every change makes a new numbered instruction set, and an agent can be put back to an
earlier one. See [[agent-teaching#instruction]].

### What is the inspector?   {#inspector}
One page that tells one piece of work to somebody who was not there: who asked, what it did step by step, who
approved what, what it cost. See [The inspector — one piece of an agent's work, readable by a person who was not there (Agents.Inspector kit)](https://osysharp.com/reference/agent/inspector/).

### What is a policy?   {#policy}
A named question about who may do something, which the app answers in one line: who uses agents, who supervises
them, who teaches them. A kit asks; the app answers. See [Who is in charge of an app's agents — every policy in one table, pausing, sponsoring, and trying work again](https://osysharp.com/reference/agent/roles/).

### What is a kit, and what is a bridge?   {#kit}
A kit is a package an app adds to gain a capability: Agents, Processes, Privacy. A bridge is a small kit that
connects two others, and arrives by itself when an app uses both. See [The map of the kits — which sits on which, and what each one is for](https://osysharp.com/reference/agent/guide-kits/).

### What is a scripted model?   {#scripted-model}
A stand-in for the AI model in a test. The test writes what the model answers; everything else is real. It costs
nothing. See [Testing an agent without a model — Says, AgentScript and Trajectory (Agents.Testing kit)](https://osysharp.com/reference/agent/testing/).

## What to read next   {#read-next}
- [Deciding what to build — four shapes of app with agents in it](https://osysharp.com/reference/agent/guide-deciding/) — the four shapes of app
- [The map of the kits — which sits on which, and what each one is for](https://osysharp.com/reference/agent/guide-kits/) — the kits, on one map
- [Agents in an Osy# app — what they are, in plain words](https://osysharp.com/reference/agent/overview/) — the guide's first page, for anybody

## Go deeper   {#see-also}
- [Agents (a class with a job, its own account, a budget and a record)](https://osysharp.com/reference/agent/index/) — every reference page about agents, by subject
- [Workflows (the run that outlives the request)](https://osysharp.com/reference/workflow/index/) — workflows
- [The security model](https://osysharp.com/reference/security/index/) — who may see and do what