Overview

Historically, getting answers from data required knowing exactly what to ask — users had to write queries and "pull" answers out of their information. This is a fundamental limitation in the enterprise, where people work in silos and don't know what they don't know.

Chief’s multi-agent system flips this model. Instead of a single AI attempting to process everything within a limited context window, Chief can deploy multiple specialized agents that work simultaneously to "push" relevant signal to the people who need it — before they even know to ask.

The platform's orchestrating agent analyzes your request, decomposes it into specialized subtasks, deploys dedicated agents to execute those tasks in parallel, and synthesizes their findings into a single, comprehensive response. Each specialized agent operates with its own 200,000+ token context window and focused toolset, enabling the system to handle research at a scale no single agent could manage alone.

Real-world example: A venture capital firm asked Chief to research all 57 of their portfolio companies simultaneously. The system deployed 57 dedicated agents — one per company — each performing 8–11 targeted searches autonomously. The result was a comprehensive report on all 57 companies, a task that previously ran out of tokens after covering just 9.

Available specialized agents

Chief includes five specialized agent types. Each is optimized for different work.

General Agent

The General agent is the default orchestrator with access to all available tools. It handles:

• Task coordination and delegation

• Multi-step workflows that need diverse capabilities

• Situations where the task doesn’t clearly map to one specialty

When to use: Default for general-purpose requests, or when the work spans multiple specialties.

💡 Tip: The General agent often delegates rather than doing everything inline—it’s the coordinator for complex workflows.

Explorer Agent

The Explorer agent specializes in web research and external information. It is optimized for:

• Internet research and fact-checking

• Reading and analyzing web pages

• Synthesizing findings from multiple online sources

• Cross-referencing claims across sites

Research methodology

• Start broad — Map the landscape with an initial pass.

• Go deep — Read primary sources fully when it matters.

• Verify — Cross-check important claims across sources.

• Synthesize — Organize results with clear attributions.

Quality standards

• Cites factual claims with source URLs where possible

• Notes confidence (e.g. well-established vs single-source)

• Calls out gaps, conflicts, or missing information

• Prefers recent sources when timeliness matters

When to use: External web research, current events, or anything not covered by your internal knowledge base.

Librarian Agent

The Librarian agent specializes in your internal document library. It excels at:

• Searching across uploaded documents

• Extracting points from spreadsheets and structured files

• Synthesizing across multiple internal sources

• Surfacing connections across documents

Search strategy

• Identify topics (finance, HR, engineering, and so on).

• Match topics to relevant assets using descriptions and metadata.

• Prefer multiple focused queries over one overly broad query.

• Revisit the same assets from different angles when needed.

Best practices

• Use natural language—not keyword stuffing.

• Break big questions into targeted searches.

• Prefer assets whose descriptions match the topic.

• Cite document sources clearly in the answer.

When to use: Deep exploration of your files, internal KB questions, or proprietary analysis.

Researcher Agent

The Researcher agent combines Explorer and Librarian. It is designed for:

• Research across both web and internal sources

• Cross-referencing external findings with internal documents

• Unified summaries with clear attribution (web vs internal)

Research strategy

• Start from internal knowledge to see what you already have.

• Use the web for updates, external context, or missing pieces.

• Cross-reference and resolve conflicts explicitly.

Quality standards

• Cites web URLs and document names

• Notes when internal and external sources disagree or complement each other

• Surfaces gaps and uncertainty

• Prefers authoritative sources when available

When to use: End-to-end research that must merge private docs and the public web.

Plan Agent

The Plan agent focuses on gathering context and producing structured execution plans for complex, multi-step tasks. It can:

• Pull context from knowledge bases and web search

• Return a structured plan with recommended next steps

• Suggest which agent types fit each step

When to use: Big, ambiguous projects where you want a plan first, or explicit step-by-step recommendations before execution.

How agent delegation works

The three-stage orchestration process

When a task warrants agent delegation, the system follows a consistent three-stage process:

Stage 1 — Master Planning & Task Decomposition The main orchestrator analyzes your request, breaks it into specialized subtasks, identifies required tools and data sources for each, and determines how many agents to deploy and of which type.

Stage 2 — Parallel Agent Deployment & Execution Each deployed agent operates as an independent entity with its own context window and specialized toolset. Agents autonomously select and execute the tools they need (web search, knowledge base queries, structured data queries, etc.) and report back incremental progress as they work. Multiple agents run simultaneously.

Stage 3 — Result Aggregation & Synthesis The main orchestrator collects results from all agents as they complete their tasks, integrates findings from all parallel research streams, and generates a comprehensive final response — delivered in whatever format you requested.

Delegation decision process

The main agent decides to delegate when:

• Task complexity — The task requires many tool calls that would consume too much of the main context window

• Specialization needed — The task clearly benefits from a specialized agent's expertise

• Parallelization opportunity — Multiple agents can work simultaneously on different aspects

• Resource management — Delegating prevents hitting execution limits in the main agent

Important: If you can accomplish the task with 1–3 direct tool calls, do NOT delegate to an agent. The overhead is not worth it. Delegation adds latency and resource overhead — reserve it for genuinely complex tasks.

Good delegation use cases

✅ Research tasks requiring multiple web searches with follow-up queries "Research the latest developments in quantum computing"

✅ Deep knowledge base exploration across many documents "Find all information about our authentication system across docs"

✅ Complex analysis requiring 4+ tool invocations "Analyze competitor pricing strategies across their websites"

✅ Parallel execution opportunities Launch multiple Explorer agents to research different topics simultaneously

✅ Planning complex multi-step workflows before execution

Using agents in the UI

Automatic vs. explicit agent selection

Chief automatically evaluates the complexity of your prompt and decides whether agents are needed — and if so, how many. This happens behind the scenes without any action required from you.

Automatic selection (default): Chief evaluates your prompt's complexity and selects the right agents autonomously. For simpler tasks, it works in single-agent mode. For deep research, multi-source analysis, or tasks that would exceed context window limits, it automatically activates and deploys the appropriate agents.

Explicit selection: When you want direct control, you can instruct Chief precisely:

Selecting specific agents in prompts

When agents are enabled, you can explicitly specify which agent to use for your task:

1. Type @ in the prompt bar to open the mention menu

2. Browse available agents: You'll see Explorer, Librarian, and Researcher listed

3. Select an agent: Click on the agent you want to use

4. Agent badge appears: The selected agent appears as a violet badge in your prompt

Available agents in the UI:

• Explorer: Web research specialist

• Librarian: Internal knowledge specialist

• Researcher: Multi-source research specialist

Example prompt with agent selection:

@Explorer Research the latest developments in quantum computing 

When you include an agent mention in your prompt, the system will prioritize using that specific agent for the task, rather than letting the main agent decide automatically.

Agent badges and visual indicators

Selected agents appear as violet badges in your prompt with:

• Robot icon: Visual indicator that this is an agent mention

• Agent name: The name of the selected agent (Explorer, Librarian, or Researcher)

• Remove option: Hover over the badge to see an X button for removing the agent selection

Removing an agent selection:

• Hover over the agent badge

• Click the X button that appears

• The agent mention is removed from your prompt

Agent configuration and limits

Tool access filtering

Each specialized agent only has access to tools relevant to its purpose:

This filtering keeps agents focused on their specialization and prevents wasted context on irrelevant tools.

Scalability and execution limits

Chief imposes no hard limit on the number of agents that can be deployed simultaneously. The system scales to the size of your task — from a single specialized agent to dozens working in parallel.

Practical limits per execution:

Agents track their resource consumption and gracefully handle budget exhaustion by returning partial results. A shared token budget of 125,000 tokens is tracked across all agents in a given turn.

Tool preferences

Agents can define tool preferences that guide their usage patterns:

• Prefer when — Conditions where a tool should be prioritized

• Use before — Tool execution order recommendations

• Avoid when — Conditions where a tool should be skipped

These preferences help agents make better tool selection decisions during execution.

Best practices for agent selection

Choosing the right agent

Delegation patterns

Sequential delegation Main agent → Plan agent (creates plan) → Execute plan steps with appropriate agents

Parallel delegation Main agent → Multiple Explorer agents (research different topics simultaneously) → Synthesize results

Hierarchical delegation Main agent → Researcher agent → Researcher delegates to Explorer and Librarian → Results flow back up

Iterative delegation Main agent → Agent completes task → Main agent analyzes results → Delegates follow-up tasks if needed

FAQ