Design
Docs
2026
ai-agents
agent-architecture
llm-systems
- Core Agent Model
-
The building blocks
The core components of an AI agent and when the agent loop is worth the overhead.
-
The System Prompt & Context Engineering
System prompts and assembled context as reliable agent behavior at runtime.
-
Structured Output & Constrained Decoding
Reliable agent outputs with schemas, constrained decoding, and validation.
-
The ReAct pattern
The loop that turns a stateless model into an agent that reasons, acts, and observes.
-
Planning & Task Decomposition
Turning broad goals into concrete plans, subtasks, and executable steps.
-
Tools & Function Calling
Exposing functions to models, validate arguments, and keeping tools use reliable.
-
Tool Servers & MCP
Discovering and calling tools across process boundaries using external tool servers and MCP.
-
Memory & Context Engineering
Conversation history management, retrieval, and context-window budgets across turns.
-
RAG as an Agent Pattern
Retrieval augmentation as an agent pattern for grounded, iterative information use.
-
Personalization, User Models & Preference Learning
Building evidence-backed user models that adapt agent behavior with scope, consent, transparency, and control.
- Execution and Control Flow
-
Workflow Orchestration
Prompt chains, routing, parallelization, workers, and evaluators.
-
State Machines & Hybrid Control Flow
Combining deterministic state transitions with flexible agent reasoning.
-
Sequential & Parallel Patterns
Deterministic multi-agent pipelines sequence, fan out, vote, and merge agent work.
-
Agent Composition & Reusable Primitives
Building composable agents with reusable orchestration primitives.
-
Model Selection & Routing
Balancing latency, cost, capability, and confidence.
-
Loop, Review-and-Critique & Iterative Refinement
Improving agent output through explicit review and feedback.
-
Dynamic Agent Orchestration
Discovering, selecting, and coordinating agent capabilities at runtime.
-
Long-Running & Durable Agents
Persisting progress, suspend, resume, and survival of long-running workflows.
-
Temporal Reasoning & Scheduling
Building agents that reason about time, deadlines, and deferred work.
- Multi-Agent Architectures
-
Multi-Agent Systems
Specialized agents with separate roles and tools.
-
Agent-to-Agent Communication
Discover, message, stream, and delegate work to other agents across system boundaries.
-
Coordinator & Hierarchical Patterns
Dynamic work delegation to specialists and results synthesis.
-
The Swarm Pattern
Peer agents collaborating, critique, and hand off work without a central coordinator.
-
Collaborative State & Conflict Resolution
Coordinating shared state safely across concurrent agent systems.
-
Agent Economics & Marketplace Design
Modeling agent costs, pricing, incentives, and marketplace dynamics.
- Advanced Reasoning and Adaptation
-
Reasoning Beyond ReAct
Reasoning strategies that extend ReAct with branching, voting, and reflection.
-
Knowledge Graphs & Structured Reasoning
Query, traverse, and reason using graphs.
-
Retrieval-Augmented Actions
Retrieving successful trajectories to guide agent actions.
-
Fine-Tuning & Distillation for Agents
Fine-tuning and distillation of agent behavior for faster, cheaper, and more consistent results.
-
Self-Improving Agents & Tool Creation
Learning from runs, refine prompts, and create tools safely.
-
Simulation & Synthetic Environments
Testing agents safely, repeatedly, and at scale.
- Specialized Agent Types
-
Coding Agents
Files, tests, terminals, and deterministic feedback.
-
Computer-Use Agents
Operating graphical software through screenshots, clicks, keystrokes, and visual feedback.
-
Multimodal Agents
Agent reasoning across text, images, audio, video, and structured media inside one loop.
-
Agentic Search & Deep Research
Iterative search, triangulating sources, and synthesizing cited answers.
-
Data Agents & SQL Generation
Natural-language database access through safe text-to-SQL agent design.
-
Voice and Real-Time Conversational Agents
Low-latency speech agents built around streaming audio and turn-taking.
-
Edge & On-Device Agents
Local agent design under privacy, latency, connectivity, and hardware constraints.
- Platform, Integration, and Operations
-
Agent Frameworks - Build vs. Buy
Trade-offs between agent frameworks and custom orchestration.
-
Embedding Agents in Existing Systems
Integrating agent capabilities into established applications and infrastructure.
-
Streaming, Speculative Execution & Latency Engineering
Reducing real and perceived latency in multi-step agent systems.
-
Error Handling & Recovery
Failures classification, retry safely, fall back, checkpoint, and recover.
-
Observability & Testing
Traces, metrics, evaluations, and regression tests as a measure of agents behavior.
-
Benchmarking & Evaluation Design
Evidence about quality, regressions, and real-world readiness.
-
Cost Optimization
Compounding agent costs across loops.
-
Agent UX & Transparency
Trustworthy agent interfaces.
-
Agent Lifecycle Management
Versioning, testing, deployment and monitoring in composite systems.
- Safety, Trust, and Governance
-
Guardrails & Safety
Runtime checks constraints and side effects beyond prompt instructions.
-
Human-in-the-Loop
Review, approval, correction, and escalation on risky actions.
-
Agent Authorization, Delegation & Consent
Carrying user authority safely through agents, sub-agents, tools, approvals, and external systems.
-
Trust Calibration & Selective Abstention
Calibrating agent confidence.
-
Agent Identity & Persona Stability
Maintaining consistent agent behavior across long-running interactions.
-
Security
Model reasoning as untrusted process containing damage with layered controls around actions.
-
Agents in Adversarial Environments
Defending agents against manipulation, injection, and adaptive attacks.
-
Compliance, Audit Trails & Regulated Agents
Regulated domains with compliance controls, explainability, and audit trails.
-
Federated & Privacy-Preserving Agents
Designing agents that respect privacy, residency, and data boundaries.
- Putting It Together
-
Putting It Together
A production reference architecture connecting agent reasoning, tools, state, trust, operations, and governance.