Agentic Blueprints 34
Filtersactive
Supervisor Team
A supervisor decides who works next — a researcher who searches or a coder who edits the repo — and assembles the result when the work is done.
Deep Research
Splits a hard question into sub-questions, researches them in parallel to keep the wall-clock low, then synthesizes one brief with citations.
Agentic RAG
Retrieval that doesn't trust its first search: it grades the results, rewrites the query when they're weak, and only answers once the context is good.
Plan and Execute
A planner breaks a goal into steps, an executor works them one at a time, and a replanner adjusts the plan as reality comes back — until it's done.
Reflection Loop
A generator drafts, a critic tears it apart, the generator revises — repeat until the critic has nothing left to flag. Quality through self-review.
Sequential Workflow (AutoGen pattern)
Four agents in a strict pipeline, each transforming the previous agent's output — the assembly-line shape, minimal and clean.
Semantic Router
A semantic router classifies each incoming message and dispatches it to the right agent over a distributed pub/sub runtime.
Reflection (AutoGen pattern)
A coder agent and a reviewer agent in a critique loop: code, review, revise, until the reviewer approves.
Negotiation Battle Simulator
Two agents negotiate against each other in a simulated deal — agent-vs-agent with a live UI, built on Google ADK and AG-UI.
Magentic-One
Microsoft's generalist five-agent team: an orchestrator directing a web surfer, file surfer, coder, and terminal to solve open-ended tasks.
Selector Group Chat
An LLM selector reads the conversation and picks which agent speaks next — planner, searcher, or analyst — instead of a fixed turn order.
Distributed Group Chat
A writer and an editor collaborate in a group chat where every agent runs in a separate distributed runtime, talking over gRPC.
Handoffs (AutoGen pattern)
Swarm-style handoffs in AutoGen: triage, sales, and refund agents transfer the conversation to whoever should own it next.
Concurrent Agents (AutoGen pattern)
Fan a task out to multiple agents at once and gather their results — AutoGen's minimal parallel pattern.
Group Chat (AutoGen pattern)
A group-chat manager coordinates role agents — writer, editor, and friends — publishing turns to a shared conversation.
Mixture of Agents
Layers of worker agents refine the previous layer's answers, with an orchestrator combining them into a stronger final response.
Self-Evolving Agent
An EvoAgentX system that generates its own multi-agent workflow for a goal, executes it, verifies the result, and repairs itself.
Multi-Agent Debate
Several solver agents answer the same problem, read each other's answers, and revise over rounds until the majority converges.
Hierarchical Agent Teams
A supervisor of supervisors: a top-level orchestrator directs a research team and a document-writing team, each with its own internal supervisor.
Self-RAG
Retrieval with self-reflection at every step: grade the documents, check the answer for hallucination, and loop until it's grounded.
Multi-Agent Collaboration
A researcher agent and a chart-generator agent divide a task and route work between themselves until the answer is a finished chart.
LLMCompiler
A planner streams a DAG of tasks, workers execute them in parallel as dependencies clear, and a joiner decides to finish or replan.