> ## Documentation Index > Fetch the complete documentation index at: https://playerzero.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Hive Mode > Deploy teams of specialist AI agents to tackle complex investigations, debugging, and research — with real-time coordination and cross-checking. ## What Is Hive Mode? Hive Mode changes how PlayerZero's AI Player approaches your problem. Instead of a single agent working through the task alone, Hive Mode turns the AI into an **orchestrator** that designs and deploys a team of specialist agents — each focused on a specific angle — and then synthesizes their findings into a unified answer. This is the right choice when accuracy matters more than speed: complex debugging, root cause analysis, deep research across multiple subsystems, or any investigation where you want multiple independent perspectives cross-checked against each other. *** ## Agent Mode vs Hive Mode | | Agent Mode | Hive Mode | | ---------------- | ----------------------------------------------------------------- | -------------------------------------------------------------------------- | | **How it works** | A single focused agent tackles your task directly using its tools | Multiple specialist agents coordinate to solve the problem in parallel | | **Speed** | Faster — one agent, one thread of work | Slower — multiple agents, but running in parallel | | **Accuracy** | Good for straightforward tasks | Higher accuracy for complex problems — cross-checking reduces blind spots | | **Best for** | Quick lookups, simple code questions, lightweight investigations | Debugging, root cause analysis, deep research, multi-system investigations | *** ## Switching Modes Toggle between Agent and Hive using the **mode slider** next to the chat input. * The slider appears on the **homepage** when creating a new thread and in the **chat input** of an existing thread * Your preference is remembered across sessions * Workflow stages can set a **preferred mode** — when you enter a stage, the mode adjusts automatically *** ## How Hive Mode Works When you send a message in Hive Mode, the AI follows a structured approach: 1. **Reconnaissance** — The orchestrator does a few quick tool calls to understand the landscape of the problem 2. **Team design** — It designs a team of specialists, wiring them into a dependency graph (DAG) so agents that depend on each other's output run in the right order 3. **Parallel execution** — All independent agents run simultaneously, each focused on their specific objective 4. **Cross-checking** — A cross-checker agent interrogates the other agents' findings, flags contradictions, and asks targeted follow-up questions 5. **Synthesis** — The orchestrator combines the validated findings into a final answer *** ## Specialist Agents Hive Mode can deploy several types of specialist agents depending on the task: * Deeply analyzes code structure, architecture, and data flows * Reads and searches across your repositories * Traces dependencies and identifies patterns * Makes precise, well-tested code changes * Follows existing patterns and conventions in your codebase * Produces minimal, surgical edits with rationale for each change * Steps through execution paths methodically * Flags implicit assumptions, missing checks, race conditions, or silent failures * Produces step-by-step traces with severity and confidence for each finding * Reviews the combined output of other agents * Identifies contradictions, gaps, and unsupported claims * Goes back to upstream agents with targeted follow-up questions to reconcile discrepancies * Creates and manages test scenarios and playlists * Builds implementation-agnostic scenarios focused on user behavior and expected outcomes * Searches the existing scenario library for relevant coverage * Specialist agents for each connected tool — Datadog, Linear, Jira, Salesforce, and others * Query logs, metrics, traces, tickets, and other external data relevant to the investigation * Only available when the corresponding integration is connected * Investigates client-side issues using browser telemetry and session data * Searches for users, retrieves browser sessions, and inspects frontend traces * Analyzes user activity timelines and correlates with frontend errors * Searches and surfaces issue statistics — frequency, affected users, and related traces * Focused on browser-side telemetry; delegates backend or infrastructure queries to other teammates * Searches and reads organizational documents uploaded to [Knowledge Bases](/features/knowledge-bases) * Uses grep-style search across knowledge groups to find relevant content * Reads full documents with line-number references for targeted analysis * Only available when at least one knowledge group exists in your organization *** ## The Hive Panel While agents are working, the **Hive tab** in the side panel shows real-time status: * Each agent appears as a node with a **status indicator** — pending, running, completed, or failed * **Live breadcrumbs** show what each agent is currently doing (e.g., "Reading auth-service/login.ts", "Searching for rate limit errors") * Animated indicators show when agents are **communicating with each other** — exchanging findings, asking follow-up questions, or sharing context * Click any agent to see its objective and dependencies Each specialist agent runs as its own **thread** in the channel. Beyond the Hive tab, these teammate threads appear in the channel's thread roster grouped by workflow, so you can open any one of them directly and follow its full history. The **coordinator** keeps the team aligned with the channel objective and relays context between threads as the work proceeds. In the chat stream, you'll see: * A **"Spawning hive..."** animation when agents are being deployed * A **status list** showing each agent's title, progress, and current activity * Completed agents appear with a checkmark; failed agents show an error indicator *** ## When to Use Each Mode | Scenario | Recommended Mode | | ---------------------------------------------------------------------------------- | ---------------- | | "What does this function do?" | Agent | | "How does authentication work across all services?" | Hive | | "Find the file that handles payment processing" | Agent | | "Why are users getting 500 errors on the checkout page?" | Hive | | "Add a log statement to this handler" | Agent | | "Investigate this production incident — check code, telemetry, and recent changes" | Hive | | "Create test scenarios for the new billing feature" | Hive | | "What changed in this PR?" | Agent | | "Is this bug a regression from a recent change?" | Hive | **Rule of thumb:** If the question has one clear answer in one place, use Agent. If it requires investigating multiple angles, comparing evidence, or cross-referencing systems, use Hive. *** ## Get Started 👉 [AI Player overview](/features/ai-player) 👉 [Setup guide](/developer-guide/quickstart-guide)