Operators

Operators are the AI workers in your guild. Each one has a role (Data Analyst, Engineer, Creative Director, Marketing Strategist), a personality, and a set of capabilities they can use to actually do work.

The Operators page lives at the top of your guild's left sidebar. Three tabs:

• Active Operators — who you've hired, shown as an org chart or list.
• Marketplace — pre-built operators you can hire in one click.
• Roles — reusable role templates (the "what they do") that operators can be assigned to.

Why "operator" and not "agent"

An agent is the underlying primitive — an LLM-driven actor with tools. An operator is an agent with a job. The distinction matters in two places:

• Hiring: you don't write a prompt, you assign a role. The role brings default instructions, tone, capabilities, and an expected scope of work.
• Operations: operators behave like employees. They escalate when stuck, log decisions, and produce traceable artifacts you can review.

In code you'll still see "agent" — operators are agents — but in the UI and docs, "operator" is the right word.

Anatomy of an operator

Operator
├─ Role               e.g. "Senior Marketing Strategist"
├─ Personality        warmth, precision, proactivity, creativity
├─ Autonomy           how often they escalate vs decide alone
├─ Skills             installable capability bundles (e.g. "SEO Auditor")
├─ Actions            tool methods they can invoke
├─ Workspace          where they read and write files
├─ Memory             accumulated context across conversations
└─ Model              the LLM that drives them (defaults from guild config)

You set most of these once when you hire. Adjust later from the operator's detail page.

The LLM behind your operators

At the bottom of the Operators page, the LLM Models block sets your guild's default model. Every operator inherits this unless you override it on the operator itself.

There are three places model behaviour can be tuned:

1. Guild default — here, on the Operators page. Sets the model used "for all operations unless specifically overridden".
2. Per-operator override — open any operator's detail page to pick a different model than the guild default.
3. BYOK keys — register your own OpenAI / Anthropic / Google keys under Usage → BYOK so calls bill against your provider account instead of Guilde credits.

Hiring an operator

Two paths:

1. Solo — one all-purpose operator, good for "I just need an assistant".
2. Team bundle — pre-built rosters (Marketing studio, Sales squad, Dev team, Support squad). Each operator in the bundle comes with role-specific defaults.

The onboarding wizard suggests a bundle based on what you said in the segmentation step. You can ignore the suggestion and pick anything.

Autonomy

Each operator has an autonomy level (1–5). Higher = more decisions made without asking, lower = escalates earlier.

• 1–2 — escalates often. Good while you're learning what an operator can do, or for high-stakes work (legal, money, public statements).
• 3 — default. Acts on the obvious, escalates the ambiguous.
• 4–5 — runs on its own. Use for repetitive, low-stakes work where speed matters more than precision.

You can change autonomy any time. It's not a contract.

Memory

Operators accumulate context inside their memory store — what they've learned about your company, your style, prior decisions. Memory is per-operator by default; an operator's memory doesn't bleed into another's.

Operators vs agents (power-user view)

The sidebar item is Operators, but the URL is /agents/active-operators. There's a deliberate separation under the hood:

• /agents is the raw agent layer — the underlying primitives.
• /operators (the Active Operators tab) is the operator layer — agents with a job, personality, autonomy.

You'll almost always live in the Operators tab. The raw /agents surface exists for power users who want to wire agents that don't fit the "employee" mental model.

What's next

• Conversations — how to talk to your operators
• Capability overview — actions, MCP, skills
• BYOK LLM keys — bring your own model provider
• Hire your first operator — full walkthrough