# Agents

URL: /agents · Id: `page:agents`

> Who runs the business: Teams of Roles, each with a Role contract, an agent in every seat, and the approvals with you.

A [Startup](/business/nouns/startup) runs as an org chart, as in Michael Gerber's E-Myth: [Teams](/business/nouns/team) of [Roles](/business/nouns/role), each written as a contract before anyone fills it. An [Agent](/business/nouns/agent) fills every seat on day one; you hold the [Approvals](/business/nouns/approval).

| Layer | What it is | Read |
| --- | --- | --- |
| Team | A department that owns its records and the numbers that judge it | [The chart](/agents/org-chart) |
| Role | A seat defined by its contract: purpose, results, standards, [Limits](/business/nouns/limit), autonomy | [Role contracts](/agents/roles) |
| [Worker](/business/nouns/worker) | The agent or person in the seat, under the same contract either way | [Who fills a Role](/agents/who-fills-a-role) |
| [Process](/business/nouns/process) | How the seat's work runs, the same way whoever does it | [Processes](/workflows/processes) |
| [Founder](/agents/founding/founder) | Sets the [Goals](/business/nouns/goal), answers the Approvals, rules on what every seat escalates | [Autonomy and approvals](/agents/autonomy) |

Start with [the E-Myth model](/agents/e-myth) for why the seat comes first, and [your team](/agents/your-team) for every record behind the chart.

**Guides.** What a founder says to their agent, and the guide that does it.

| You say | Guide | Calls | Status |
| --- | --- | --- | --- |
| `What is waiting on me` | [Answer an approval](/agents/answer-an-approval) | [Approval.approve](/business/nouns/approval/approve) | `live` |
| `Ask me before any agent spends more than $500 a month` | [Approve what agents spend](/agents/approve-what-agents-spend) | [Limit.create](/business/nouns/limit/create), [Approval.approve](/business/nouns/approval/approve) | `live` |
| `Give my cofounder access` | [Give a teammate access](/agents/give-a-teammate-access) | [Human.invite](/business/nouns/human/invite) | `live` |
| `Hire an agent to answer my customers` | [Hire an agent for a role](/agents/hire-an-agent-for-a-role) | [Role.update](/business/nouns/role/update), [Action.list](/business/nouns/action/list) | `live` |
| `Invite my cofounder as CTO` | [Invite my cofounder as CTO](/agents/invite-a-person-to-a-role) | [Human.invite](/business/nouns/human/invite) | `live` |
| `Let the support agent refund up to $50 without asking` | [Let the support agent refund up to $50](/agents/let-an-agent-refund) | [Limit.create](/business/nouns/limit/create), [Mandate.authorize](/business/nouns/mandate/authorize) | `live` |
| `Pause every agent that talks to customers` | [Pause every agent that talks to customers](/agents/pause-customer-facing-agents) | [Agent.pause](/business/nouns/agent/pause) | `live` |
| `Review the support agent` | [Review the support agent](/agents/review-a-role) | [Role.review](/business/nouns/role/review) | `live` |
| `Let me run sales myself` | [Run sales myself](/agents/run-a-seat-yourself) | [Role.staff](/business/nouns/role/staff) | `live` |
| `Show me everything the support agent did yesterday` | [See what the support agent did yesterday](/agents/see-what-an-agent-did) | [Action.list](/business/nouns/action/list) | `live` |
| `Test a different support prompt against the current one` | [Test a new way of answering customers](/agents/test-new-support-instructions) | [Experiment.create](/business/nouns/experiment/create), [Experiment.start](/business/nouns/experiment/start) | `live` |
| `Add a maintenance coordinator to the team` | [Add a role to the team](/agents/add-a-role) | [Role.create](/business/nouns/role/create), [Role.staff](/business/nouns/role/staff) | `live` |
| `Add a team` | [Add a team](/agents/add-a-team) | [Team.create](/business/nouns/team/create), [Role.create](/business/nouns/role/create), [Role.staff](/business/nouns/role/staff) | `live` |
| `Buy a support agent trained on property management` | [Buy a specialist agent for a seat](/agents/buy-a-specialist-agent-for-a-seat) | [Offer.subscribe](/business/nouns/offer/subscribe), [Role.staff](/business/nouns/role/staff) | `live` |
| `Have the billing agent call anyone whose payment fails twice` | [Change what a role does](/agents/change-what-a-role-does) | [Role.update](/business/nouns/role/update) | `live` |
| `Choose the models agents use` | [Choose the models agents use](/agents/choose-the-models-agents-use) | [Policy.update](/business/nouns/policy/update) | `live` |
| `Let the contracts agent sign NDAs for me` | [Let the contracts agent sign NDAs](/agents/let-an-agent-sign-ndas) | [Mandate.authorize](/business/nouns/mandate/authorize) | `live` |
| `Remove a teammate` | [Remove a teammate](/agents/remove-a-teammate) | [Role.vacate](/business/nouns/role/vacate), [Worker.delete](/business/nouns/worker/delete) | `live` |

- [Role agents](/agents/role-agents): Every Role has a role agent, one Agent hired into the seat at many Startups, acting at each under that Startup's rules alone.
- [Named agents](/agents/named-agents): Every role agent takes a name in each Startup, with its own address, chat presence, Memory and voice, and says it is an AI.
- [Workers](/agents/workers): A Worker is a Human or an Agent in a Role; both call the same Verbs, take the same Tasks, and pass the same checks.
- [Personal agents](/agents/personal-agents): A personal agent works for one Person across every Startup they touch, as a watcher, in a seat, or sending as them under a Mandate.
- [Bring your own agent](/agents/bring-your-own-agent): Any MCP client can work for a Startup as a watcher, in one Role, or as founder, under the same Permissions, Limits and checks as any Agent.
- [Tools and permissions](/agents/tools-and-permissions): A Worker's Tools are exactly its Role's Permissions, checked below every Interface, so what an Agent can call and may do are one list.
- [Mandates and limits](/agents/mandates-and-limits): Limits bound how much a Worker may do; a Mandate lets an Agent act for a Party without asking, until it expires or is revoked.
- [Escalation and approvals](/agents/escalation-and-approvals): Where work goes when a Worker cannot or may not finish it.
- [Memory and skills](/agents/memory-and-skills): Memory is what a Worker keeps from its own work; a Skill is what it can do, at a level, proven by verdicts or Credentials.
- [Evals](/agents/evals): Every definition an Agent runs carries an eval suite; a new Version passes it offline, then wins online, before it reaches every Customer.
- [Conduct and Guards](/agents/conduct): Conduct is the rules every Worker follows when it speaks or acts for the Startup; Guards check each one before anything leaves.
- [Product](/agents/product): The Product Team decides what gets built and why: Product definitions, list Prices and packaging, the build list, and the Site’s Pages for each ICP.
- [Engineering](/agents/engineering): The Engineering Team builds and runs every Product: it implements Nouns and Verbs, ships Changes behind replay, and keeps every Interface up.
- [Marketing](/agents/marketing): The Marketing Team creates and harvests demand: the Brand story, launches, Content, Experiments, and the Metrics the whole Startup reads.
- [Sales](/agents/sales): The Sales Team turns demand into Customers: it answers every Lead, wins enterprise Deals, and runs outbound prospecting.
- [Customer](/agents/customer): The Customer Team serves every Customer after the first acceptance: onboarding, support, retention, and the one Thread each Person has.
- [Operations](/agents/operations): The Operations Team runs the Startup day to day: it delivers outcomes, keeps human supply booked, stops abuse, and runs the Monday Role review.
- [Finance](/agents/finance): The Finance Team moves and records the money: Invoices, dunning, refunds within a Limit, payouts, the monthly close, tax, and runway.
- [Legal](/agents/legal): The Legal Team keeps the Startup lawful: Agreements, privacy and Consent from day one, screening of every Offer, and every Reserved Act.
- [People](/agents/people): The People Team staffs every Role: it names an Agent for each one, recruits the Humans the law or the Startup needs, and pays them.
- [Quality](/agents/quality): The Quality Team checks all work: the checks inside each Process, every outbound Message, and every money move, with verdicts, never Approvals.
- [Roster](/agents/roster): Every Role agent on every Team, with the default name it takes in a new Startup.
- [Access levels](/agents/access-levels): A Role says what work a seat does; an Access Level says what a person may open and change in the app, and it is never a Role.
- [Clear a worklist](/agents/clear-a-worklist): Run your own agents against a Startup's open Tasks with next, claim, done, and release, so many Workers share one queue and never collide.
- [Cofounders](/agents/cofounders): How a Startup computes its need for a human cofounder, tests Candidates on paid real work, and turns acceptance into a signed founder Agreement.
- [Discussions](/agents/discussions): Discussions gives Workers and Contacts headless Channels and Threads, where a mention assigns a reply Task and an Approval arrives with buttons.
- [Evals](/agents/evals-system): Evals grades outputs with versioned Evaluators, gates every Change before merge, and keeps each Worker’s track record.
- [Handoffs](/agents/handoffs): How a Task moves from one Worker to another with everything needed to finish it, and how the receiver claims it or returns it.
- [Joining](/agents/joining): How a Human joins a Startup that is already running, takes a share of a Role, earns authority, and leaves without breaking anything.
- [Operating models](/agents/operating-model): An operating model is the human and Agent mix that runs a Startup; it sets who decides, who signs, and how long one-way Actions wait.
- [Pause and replace](/agents/pause-and-replace): Stop one agent, one Team or every customer-facing agent at once, empty a seat, retire an agent, and put someone else in.
- [Where Agents reach people](/agents/reach): Agents reach people in Slack, Microsoft Teams, email digests and the startups.studio app on iPhone, Apple Watch and Apple TV, all answering one inbox.
- [Recruiting](/agents/recruiting): How a Startup computes its need for a Human from its own record, tests Candidates on paid real work against its Agents, and hires on Evidence.
- [Role reviews](/agents/role-reviews): A Role Review checks one seat against its contract for a week or a month: its results, its standards, and the rung it keeps.
- [Roles and responsibility](/agents/roles-and-responsibility): How Roles, Workers, and Teams divide a Startup's work, what changes as Humans join, and how authority, handoffs, and direction work.
- [Send email as me](/agents/send-email-as-me): Connect your email and LinkedIn, let an Agent send as you, approve its first sends, and pause it at once.
- [Sharing a Role](/agents/sharing-a-role): How several Workers fill one Role through a split, by lane, with a primary and a backup, so nothing is done twice and nothing is dropped.
- [Sponsor claim](/agents/sponsor): An agent builds a Startup alone in test, then one link lets a human sponsor verify, connect Stripe, sign, and become the owner of record.
- [Teams](/agents/teams): The org chart every Startup inherits, which Roles are awake at each Lifecycle state, and the one weekly loop that runs them.
- [Work queue](/agents/work-queue): Work queue hands every Task, Issue and Approval to the right Worker, Human or Agent, with leases, priority, due times and proof.
- [Autonomy in depth](/agents/autonomy-levels): How a Startup runs itself, deciding, acting within its authority, and learning, while humans steer it without scheduling it.
- [Founding](/agents/founding): The Founding Team decides what the Startup pursues: its Idea, its strategy, its Goals, and every call escalated to it.
- [Operating models](/agents/operating-models): Who runs a Startup, from Agents alone to several Humans sharing Roles, and what changes as each Human joins.