team-of-agents

Hire 17 domain experts in one install. Invoke any specialist directly,
or let the orchestrator plan, dispatch, and synthesise the work for you.

Install · How It Works · Orchestrator · Specialists · Examples · Contributing · Changelog

See it in action

One request. Multiple specialists dispatched in parallel. One synthesised answer.

/orchestrator   my checkout conversion is dropping, help me figure out why and fix it

→ Identified: UX Researcher + Data Analyst + Frontend Reviewer  [parallel]
→ UX Researcher    [High]  users don't trust the payment form
→ Data Analyst     [High]  mobile abandonment is 2× desktop
→ Frontend Reviewer [High] submit button is below the fold on mobile

→ Chaining findings → Backend Engineer + Frontend Designer  [parallel]
→ Senior Engineer review pass...
→ Synthesis complete. Full trace log attached.

What it does

• Orchestrates multi-role tasks automatically, drop in a brief and the orchestrator decomposes it, assigns specialists, runs them in parallel where possible, and assembles one answer
• 17 specialists across the full SDLC, backend, frontend, Kotlin, SRE, DevEx, QA, product, project, UX, data, SEO, document writer, and more
• Parallel execution built-in, independent tasks run simultaneously; sequential tasks chain findings forward automatically
• Confidence signals on every output, every specialist returns High / Medium / Low so the orchestrator and you know what needs a second look
• No silent changes, every specialist announces what it will do, waits for your approval before writing files or running commands, and reports what was done
• Trust tiers, T1 (research), T2 (file writes), T3 (commands), declared in the plan so you always know what requires your sign-off

Install

/plugin marketplace add pranav8494/team-of-agents
/plugin install team-of-agents@team-of-agents

How It Works

Think of this plugin as a consultant firm. The orchestrator is the project manager, it listens, breaks the problem into pieces, and hands each piece to the right expert. Each expert does their slice and hands the work back. The project manager assembles it into one answer.

The three layers

You describe a task in plain English.

The orchestrator reads your request, identifies which specialists are needed, shows you the plan, waits for approval, dispatches each specialist, collects their answers, and produces one combined result.

The specialists (17 of them) each handle one domain, backend, frontend, data, UX, SRE, and so on. Each does its slice and nothing more.

What happens step by step

You type a request
       ↓
Orchestrator reads it → breaks it into subtasks → shows you the plan
       ↓
You confirm → orchestrator dispatches specialists
(independent tasks run in parallel; dependent tasks run sequentially)
       ↓
Each specialist returns output + a confidence signal (High / Medium / Low)
       ↓
Orchestrator checks confidence:
  High   → use it directly
  Medium → flag the assumption, ask you to confirm before acting
  Low    → surface the gap, re-dispatch with more context
  BLOCKED → ask you what's missing, then retry
       ↓
Optional: senior-engineer reviews all outputs for errors or conflicts
       ↓
Orchestrator assembles everything into one answer with a full trace log