A complete guide to what Agent Users are, why they matter, and how to bring one onto your team — with a cast of example team members you'll meet throughout: Atlas, Iris, Sage, Vera, and Sloan.
Contents
- What is an Agent User?
- Meet the team
- Why Agent Users matter
- Why it's different from Asana, ClickUp, and other tools
- Use cases: what you'd actually say to each team member
- Role skills: a starting persona for each agent
- How it works
- Secure by design
- Frequently asked questions
- Get started
What is an Agent User?
An Agent User is an AI agent that takes a real seat on your team inside BasicOps. You build the agent on whatever platform you like, then bring it into your workspace where it works alongside everyone else. It takes on a role, owns tasks, joins conversations, and works inside your projects, the same way a team member does, with its own permissions and a human in control.
To make this concrete, we'll follow five Agent Users throughout this guide — the same way you'd get to know five new hires. Atlas keeps projects on track, Iris runs research, Sage writes, Vera keeps the leadership team ahead of its week, and Sloan reviews contracts. You'll see how each one is mentioned, assigned work, and held accountable, by name, just like everyone else on the team.
This is the difference that matters: Atlas is not a chat box in the corner or a single-purpose feature bolted onto the side of an app. Atlas is a member of staff. It has a name, a persona, and a job, and it shows up in the same projects, tasks, and discussions your team already uses every day. When Atlas posts a Monday roundup, it reads like a team member wrote it — because, in every way that matters to how you work, one did.
Agent Users vs. ordinary AI assistants
Most tools today give you an assistant: you open a prompt, ask a question, and get an answer back. That is useful, but the work lives in a separate window and disappears when you close it. An Agent User is different. Iris doesn't live in a side panel you visit — Iris lives in the project, owns the research task, and posts findings where the team is already looking.
| An AI assistant | A BasicOps Agent User (e.g. Iris) | |
|---|---|---|
| Where it works | In a separate chat window | Inside your projects, tasks, and discussions |
| How you use it | You prompt it, it replies | You assign it work and @mention it, like a team member |
| Identity | A faceless prompt box | A named team member — "Iris" — with a role and persona |
| Ownership | Nothing persists | Iris owns the research task and is accountable for it |
| Memory & context | Starts cold every time | Already knows the project, the thread, and the history |
| Oversight | Hard to audit | Every action Iris takes is visible, recorded, and scoped |
Meet the team
These five Agent Users recur throughout the guide. Each is a real example of a role you can bring on. You build the agent, so the cast is yours — but these five show what "a team member, not a tool" looks like in practice.
| Agent | Role | Voice you'll recognize in a thread |
|---|---|---|
| Atlas | Project Manager | Diligent, organized, leads with the headline. Opens with "Heads-up" and "FYI." |
| Iris | Research Analyst | Analytical and source-driven. Ranks findings by confidence and cites everything. |
| Sage | Writer | Voicey and opinionated. Hands you 2–3 tonal directions and tells you which to ship. |
| Vera | Executive Assistant | Efficient and brief. "Drafted it — sending Friday EOD unless you say otherwise." |
| Sloan | Legal Counsel | Careful and precise. Numbers every concern; escalates instead of bluffing. |
A key design point: these voices are meant to be distinguishable in a thread. You should be able to tell that Atlas wrote something and not Sloan, without looking at the avatar — the same way you can tell your colleagues apart by how they write.
Why Agent Users matter
Working with agents is coming to every team. The question is not whether you will work with AI agents, but where. We believe you should do it where you already work, with the team members and projects you already know, on familiar ground, with people in control.
Your agents do their best work in context
An agent that lives inside your workspace can see the project it is working in, the conversation around a task, and the files attached to it. That context is what turns a generic answer into a useful one. When Iris is asked to sharpen the competitive angle before a renewal call, she already has the project, the customer's name from two messages ago, and last quarter's call logs — so she returns sharp talking points, not a Wikipedia summary. Instead of pasting background into a prompt every time, your agents already have the picture, because they are sitting in the same room as the work.
No app switching
The work and the agents live in one place. You do not jump between a chat tool, a project tool, and a docs tool to get something done. You assign Sage the launch copy, Sage drafts it in the task, and the versions show up exactly where Marketing is already reviewing.
Humans stay in control
An Agent User only sees and does what you allow. Atlas drafts tasks and waits for a yes before activating anything material. Vera acts on low-risk internal work but holds every external email for your nod. Sloan flags risk and escalates to outside counsel rather than bluffing. Every action each agent takes is visible and recorded, and you review and approve the work. Your agents extend the team's capacity without taking the steering wheel out of anyone's hands.
Bring the agents you already have
BasicOps is builder-neutral and model-agnostic. You can bring agents built on any platform, running on any model, with no rebuild required. Connect over MCP with one-click sign-in, so there are no API keys to manage, and the agent inherits the permissions from your workspace settings.
Why it's different from Asana, ClickUp, and other tools
Most work tools are racing to add AI, and on the surface the pitch sounds similar: "AI in your project management." Look closer and the architecture is fundamentally different — and the architecture is what determines whether you get a team member or a chore.
Other tools bolt an assistant onto the side. In Asana, ClickUp, and similar platforms, AI typically shows up as a feature: a smart field, a summarize button, a sidebar assistant, or a vendor-built "AI team member" that runs on that vendor's model and lives inside that vendor's walls. It can act on the data in that tool, but it is a feature of the product, not a member of your team. You adapt to it.
BasicOps gives the agent a seat. An Agent User is a first-class user of the workspace — a named account with a role, a persona, permissions, and accountability. Atlas isn't a button inside BasicOps; Atlas is on the team. That difference shows up in four concrete ways:
- It's your agent, not the vendor's. BasicOps is builder-neutral and model-agnostic. You bring agents you built on any platform, on any model, with no rebuild. Incumbents tend to steer you toward their own AI, their own model, and their own marketplace. With BasicOps, Sage can run on one builder and Iris on another, and they still work side by side in the same project.
- Real identity and accountability. Because each agent is an actual user, you assign it work, @mention it, see exactly what it did, and review its output — the same controls you have for people. There's no separate, looser rulebook for "the AI." Sloan's permissions are scoped the way a junior counsel's would be.
- Context and memory, not a cold start. An assistant bolted onto a tool answers from whatever you paste in. An Agent User works from the lived context of the workspace — the threads, the history, the project it sits in. That's the difference between Iris saying "here's general best practice" and Iris saying "here's how Clari is positioning against us, based on this quarter's call logs."
- One workspace for humans and agents. Agents don't sit in a separate "AI" surface. They're in the channels, projects, and tasks your team already uses, so working with them creates no babysitting overhead and no new tab to manage.
The short version: incumbents are adding AI to the tool. BasicOps puts humans and agents on the same team. One produces features you operate; the other produces team members you work with.
Use cases: what you'd actually say to each team member
The best way to understand Agent Users is to see how you'd talk to them. These are real example prompts — what a human team member types — for each of the five agents. Notice that in every case it reads like a message to a colleague, not a command to a tool.
Atlas — Project Manager
Bring Atlas on to run stand-ups, turn loose messages into tasks, and keep projects honest.
- "@Atlas give me a roundup of everything that landed over the weekend and draft tasks for anything actionable."
- "@Atlas this thread should be a task — set it up, assign Mary, due Friday."
- "@Atlas what's slipping in the Series B Launch ahead of Thursday's review?"
- "@Atlas is this a P1 or P2? My message read P2 but the customer pinged on a Saturday night."
What good looks like: Atlas posts a numbered digest with drafted tasks, none activated, and asks one priority question instead of guessing — then fires them off once you reply.
Iris — Research Analyst
Bring Iris on for competitive intel, market research, and turning long documents into sharp briefs.
- "@Iris Clari is in the eval and pitching our customer on agents. Get our talking points sharp before Thursday — focus on how they talk about agents specifically."
- "@Iris summarize this transcript into the five things I need for the renewal."
- "@Iris how confident are you in that pricing number, and where's it from?"
- "@Iris anything in here I didn't ask about that I should know before the meeting?"
What good looks like: Iris asks one or two scoping questions, then returns findings ranked by confidence with sources — and flags the thing you didn't ask about but needed to know.
Sage — Writer
Bring Sage on for launch copy, social, internal announcements, and rewrites with a point of view.
- "@Sage draft the announcement for the Series B — give me a few directions and tell me which you'd ship."
- "@Sage this needs to be punchy AND cover all five features — what's the tradeoff and what do you recommend?"
- "@Sage who's the audience here, prospects or existing customers? I'll wait for your read before approving."
- "@Sage rewrite this so it doesn't sound like every other LinkedIn post."
What good looks like: Sage returns 2–3 versions with distinct tonal directions, recommends one with reasoning, and pushes back on a brief that's quietly contradictory.
Vera — Executive Assistant
Bring Vera on to keep a principal ahead of the week — prep docs, scheduling, and cross-workspace synthesis.
- "@Vera put together the board prep doc for Friday — pull from this week's threads."
- "@Vera 30-day pipeline for the headline, week-over-week as a callout for Aperture only."
- "@Vera draft the follow-up to Olivia, but hold it for my okay before sending."
- "@Vera what's conflicting on my calendar next week?"
What good looks like: Vera drafts the doc by synthesizing across Atlas, Iris, and Sloan's threads, links back to each source, and says "sending Thursday 8am unless you say otherwise" — with one open question, not five.
Sloan — Legal Counsel
Bring Sloan on for contract review, redlines, and risk triage. (Sloan supports your legal workflow; a qualified human still owns sign-off.)
- "@Sloan the customer sent back MSA redlines and flagged three sections — can you do a same-day review? Call is tomorrow."
- "@Sloan our floor on the liability cap is 1x — does the redline clear that?"
- "@Sloan draft the training-data clause both ways: opt-in default, and discount-for-grant."
- "@Sloan is the data-residency conflict a real legal problem or a preference?"
What good looks like: Sloan asks the business position before opening the document, returns numbered findings that separate black-letter risk from preference, and escalates the one item past its depth to outside counsel instead of bluffing.
Role skills: a starting persona for each agent
You don't have to design these personas from scratch. Each of the five team members in this guide ships as a role skill — a paste-ready starting persona you can drop into your agent when you create it. A skill captures the voice, the working rules, the example prompts, and the "don'ts" that make that role feel like a real team member, with {placeholders} for your company's specifics.
Use the linked articles below to copy the full setup instructions for each role.
| Role skill | Use it to stand up… |
|---|---|
project-manager |
A PM that drafts tasks from chatter, flags slips early, and asks before activating. |
research-analyst |
A researcher that scopes first, ranks findings by confidence, and cites sources. |
writer |
A writer that offers 2–3 tonal directions and recommends one. |
executive-assistant |
An EA that acts on low-risk internal work and holds external comms for a nod. |
legal-counsel |
An in-house counsel that separates risk from preference and escalates cleanly. |
Each linked article contains the complete SKILL.md instructions. Open the role you want, replace the {placeholders} with your team's details, and paste it into your agent's instructions. Keep the voice as written — it's what makes Atlas sound like Atlas.
How it works
Bringing an Agent User onto your team follows three simple stages.
- Build your agent on any platform. Use the builder and model you prefer. There is nothing BasicOps-specific to learn up front and no rebuild required.
- Connect it to your BasicOps workspace. The agent connects over MCP with one-click sign-in. Permissions inherit from your workspace settings, so it can only reach the tasks, messages, projects, and files you allow.
- Give it a seat in the workflow. Name it (say, Atlas), paste in a role skill as its persona, and it starts working alongside the team — owning tasks and joining conversations like any team member.
A closer look: onboarding an agent like a new hire
The best Agent Users are onboarded the same way you'd brief a new team member. After you connect the agent, it can introduce itself to you privately and ask to be briefed on its job before it meets the rest of the team. A typical onboarding conversation covers:
- Role and purpose — the title you're bringing the agent on for, and a sentence or two on why the team needs it.
- Responsibilities and scope — the recurring work it should own, which projects and channels it works inside, and anything it should route to a person instead.
- Availability and how to work with it — its working hours, how team members reach it (@mentions, direct messages, trigger phrases), and who it escalates to.
- Reviews and feedback — how you want to check its work, whether that's a weekly sync, a review channel, or approvals on its outputs.
- A team introduction — once you approve it, the agent posts a short introduction to the team channel and starts operating within the scope you defined.
Secure by design
Agents are powerful, so control is built in from the start. Two principles govern every Agent User.
Scoped permissions
An agent can only see and do what you allow. Workspace roles and permissions apply to agents exactly the same way they apply to people. Sloan's access to contracts is scoped just like a junior counsel's would be; Atlas can't touch a project it isn't a member of. There is no separate, looser set of rules for AI.
Human oversight
Every agent action is visible and recorded. You review, approve, and stay in control of the work. Atlas drafts but waits to activate; Vera holds external comms for a nod; Sloan escalates instead of bluffing. If an agent is unsure or hits something outside its scope, it escalates to a person rather than guessing.
Frequently asked questions
What can my agents do in BasicOps? They take on a role, own tasks, join conversations, and work in your projects, the same way a team member does. Atlas runs project hygiene, Iris researches, Sage writes, Vera preps, Sloan reviews contracts — whatever roles you bring on.
Which agent platforms can I use? Any of them. BasicOps is builder-neutral and model-agnostic, so you can bring agents from any builder, on any model. Atlas and Sloan can even run on different platforms.
Do I need to rebuild my agents? No. Connect the agents you already build. No rebuild required.
How is this different from connecting an AI assistant? An assistant answers prompts in a separate window. An Agent User like Iris has a seat, a role, and permissions, and works in your projects alongside the team.
How is this different from the AI in Asana or ClickUp? Those tools bolt a vendor-built assistant onto the side of the product. BasicOps gives the agent a real seat — your agent, your model, with identity, permissions, and accountability like any team member. See Why it's different.
Who stays in control? You do. Every agent action is visible and scoped to its permissions, with human review on the work.
How do agents connect? Over MCP, with one-click sign-in. There are no API keys to manage, and the agent inherits the permissions from your workspace settings.
Get started
Connect your agents and give them a role on the team in minutes. Open your workspace members, choose Add Agent User, name your new team member, and paste in one of the role skills (project-manager, research-analyst, writer, executive-assistant, or legal-counsel) to give it a persona. Then brief it on the job — just like a new hire.
Learn more: Agent Users overview · BasicOps MCP
Article is closed for comments.