Subagents

Multi-agent orchestration for delegating complex, specialized workloads

What are Subagents?

Instead of relying on a single monolithic agent to handle every aspect of a complicated objective, Opulent utilizes a Subagent Architecture.

When you give Opulent a complex task, the primary agent acts as an Orchestrator. It breaks down the problem, plans the execution, and then dynamically spins up specialized "subagents" to handle the individual components. These subagents act in isolation, focusing entirely on their narrow tasks, and return their results to the orchestrator for synthesis.

Why Subagents Matter

Asking one agent to write code, design a UI, research competitors, and format a presentation in a single pass leads to context confusion and poor results. Subagents solve this:

• Specialization: Subagents can be strictly prompted for specific roles (e.g., an "Explorer" subagent built just for web scraping, or a "Coder" subagent built strictly for resolving typescript errors).
• Parallel Execution: The orchestrator can spin up 10 subagents simultaneously to research 10 different companies, turning 2 hours of sequential work into 5 minutes of parallel work.
• Context Isolation: A subagent researching competitors doesn't need the codebase context in its window, keeping its token usage low and its reasoning sharp.

How It Works

1. Decomposition and Planning

The Orchestrator agent analyzes your request. If the task is simple, it may handle it directly. If complex, it creates an execution graph.

2. Subagent Spawning

The Orchestrator spawns specialized subagents and delegates the sub-tasks. Each subagent is given:

• A specific, narrow objective.
• Only the precise context it needs.
• Access to specific tools (a web-scraper subagent doesn't need file-writing tools).

3. Execution & Verification

Subagents execute their tasks. They can autonomously verify their own work (e.g., a coder subagent running a linter) before returning the result.

4. Synthesis

The Orchestrator collects the outputs from all subagents, resolves any conflicts, synthesizes the final result, and presents it to you.

Example Subagent Topologies

Software Development Topology

"Build a new authentication flow for the frontend."

1. Orchestrator: Plans the changes needed.
2. Researcher Subagent: Reads the UI component library docs and existing auth state.
3. Coder Subagent: Writes the new React components in the sandbox.
4. Verifier Subagent: Runs the test suite and attempts to log in using the Cloud Browser Infrastructure, reporting any visual bugs back to the Coder.

Competitive Intelligence Topology

"Compile a feature comparison of our top 5 competitors."

1. Orchestrator: Identifies the 5 competitors.
2. Explorer Subagents (x5): Spins up 5 parallel subagents. Each navigates to one competitor's website, extracts the pricing and feature lists, and structures the data.
3. Orchestrator: Takes all 5 structured JSON objects, compares them, and generates the final matrix.

Managing Subagents

While subagents operate autonomously, you maintain full visibility.

• Live Monitoring: The Opulent dashboard shows the active orchestrator plan and the real-time status of all running subagents.
• Granular Control: If a specific subagent struggles or asks for clarification, you can interact with it directly without derailing the other subagents.
• Resource Limits: Administrators can set limits on how many subagents an orchestrator can spawn simultaneously to manage computational budgets.