Skip to main content

​
Why this matters

Most AI tools give you a single thread of thinking, no matter how big the question. Chief doesn’t. For a simple ask, you get a fast, direct answer. For a hard one — “compare our pricing to the three competitors in this folder and tell me where we’re exposed” — Chief quietly assembles a team of specialist agents that work in parallel, then hands you one synthesized, cited answer. Understanding this is the difference between treating Chief like a chatbot and treating it like an analyst bench that scales to the size of the problem.

​
How it actually works

Chief agents - the specialist agents with their descriptions
Every conversation starts with the General Agent, Chief’s orchestrator. It reads your request, judges how complex it is, and makes a call: handle it directly, or break it apart. You never choose — the calibration happens automatically. When a request is complex enough to overwhelm a single line of reasoning, the Planner Agent steps in first. It decomposes the task into focused subtasks and decides how many specialists to deploy and what each should focus on. From there, Chief spins up as many agents as the job needs, each working simultaneously on a different facet of the problem:
  • Explorer — the external research specialist. It scans the open web, news, research papers, GitHub, social posts, personal sites, and financial filings. You can summon it directly by typing @explorer in your prompt.
  • Librarian — your institutional memory made searchable. It works only inside your uploaded documents, project files, and transcripts, surfacing the exact passage with a citation back to the source. Summon it with @librarian.
  • Researcher — the synthesis engine and a superset of the other two. It draws from your internal knowledge and the open web at once, ideal for “how do our numbers compare to the market” questions.
When the specialists report back, the General Agent stitches their findings into one coherent, source-attributed response — visible step by step in the activity feed. You never manage the coordination; it’s invisible by design.
ℹ️ Good to Know: Agents run under Deep Intelligence mode, a paid-plan feature. If you add an agent like @explorer to a prompt, Chief’s LLM Router automatically switches that prompt into Deep Intelligence mode for you.

​
A worked example

Chief agents - the activity feed showing the agentic steps behind an answer
You ask: “Pull every churn-related complaint from last quarter’s call transcripts, then check whether our competitors shipped anything that would make those complaints worse.” A single chatbot would either read your transcripts or search the web — not both, and not deeply. Chief’s Planner splits the job: a Librarian agent mines your transcripts for churn signals while an Explorer agent scans competitor release notes and news. The Researcher reconciles the two, and the General Agent returns one brief — internal pain points lined up against external moves — with citations to both the transcript lines and the competitor sources.

​
FAQ

No. Simple, direct questions are answered directly by the General Agent in a single pass. Multi-agent orchestration only kicks in when the request is complex enough to benefit from decomposition and parallel work.
The most common reason is that you weren’t in Deep Intelligence mode. Agents are a Deep Intelligence (paid-plan) capability. Typing an agent mention like @explorer nudges the LLM Router to switch that prompt into Deep Intelligence mode; without it, straightforward prompts may be handled directly.
You can request specific specialists directly with @explorer (external research) or @librarian (your internal documents). The Planner and General Agent handle the rest of the orchestration automatically.