Agentic Craft
Build agents that survive production.
Architectures, evaluation, security, and runtime systems for reliable AI agents — implemented with HazelJS.
Choose your path
Four clear entries. Pick where you are and keep moving.
Agent OS at a glance
A layered model for operating agents as managed software — not a thin Kubernetes wrapper, and not a demo chatbot.
Control Plane
Registry, deployments, reconciliation, policy, observation
Agent DNA
Identity, model, skills, policy, permissions, SLOs
Agent Runtime
Plan, execute, critique, validate
Skill Gateway
APIs converted into governed agent capabilities
Durable Kernel
Runs, checkpoints, HITL, leases, state, recovery
Application / Backend
Business logic and integrations
Featured learning tracks
Agent Foundations
Beginner foundations on HazelJS: what agentic AI means, how AgentRuntime differs from chat and workflows, when not to use an agent, and the runtime mental model.
Agentic Development
Beginner end-to-end HazelJS course: scaffold Ops Desk, tool loop, budgets, context hygiene, HITL writes, golden tests, observability, CLI, and go-live checklist.
Context Engineering
Assemble prompts, RAG, and truncations in HazelJS so AgentRuntime sees signal—not noise.
Production patterns
Reusable designs with forces, failure modes, and trade-offs.
Augmented LLM
A single HazelJS AgentRuntime call enriched with retrieval, bounded tools, or structured output—without an open-ended agent loop.
Tool-Using Agent
A bounded AgentRuntime observe→decide→act loop where HazelJS validates tool calls via ToolRegistry and feeds observations back to the model.
Durable Agent Run
Persist AgentRun checkpoints and leases with createDurableRunStore / createSqlDurableRunStore so HazelJS agents survive crashes without double-applying side effects.
Idempotent Tool Execution
Make AgentRuntime write tools safe under retries and durable resume using idempotency keys and dedupe semantics in adapters.
Compensating Action
Undo or neutralize partial side effects after agent or workflow failure using saga/flow compensation and idempotent reverse tools.
Bounded Autonomy
Cap what a HazelJS agent may do unsupervised using capabilities, PolicyEngine effects, maxSteps, and HITL—not prompt instructions alone.
Latest practical guides
Build Your First Tool-Using Agent
Hands-on HazelJS: install packages, define @Agent/@Tool, wire AgentModule or AgentRuntime, execute with budgets, and add describeAgent tests.
From Agent Demo to Production
Promote a working HazelJS demo through Agent OS: policies, HITL, durable store, contracts, recovery, governance, Inspector, and DNA.
Design an Agent Manifest
Capture HazelJS Agent DNA: exportAgentDna, bootstrapRuntimeFromDna, CLI hazel agent run, and a reviewable YAML/JSON manifest aligned with identity, skills, policy, and SLOs.
From demo to production
A compact preview — expand it in the learning track.
- 01Define stop conditions and budgets before autonomy
- 02Validate every tool call; treat outputs as untrusted
- 03Add human approval for irreversible side effects
- 04Checkpoint run state for crash recovery
- 05Evaluate quality offline before expanding tool scope
- 06Emit traces for decisions, tools, cost, and latency
Learn by building on HazelJS
Lessons, patterns, and guides use the HazelJS Agent OS stack — @hazeljs/agent, Skillgate, MCP, memory, observability, and related packages — so you ship production agents on one TypeScript platform.
Agentic Craft is the educational field guide for building production agents with HazelJS. Examples, patterns, and learning tracks use the HazelJS Agent OS stack (@hazeljs/agent, Skillgate, MCP, memory, observability, and related packages).
Weekly production signal
One HazelJS production agent pattern, one teardown, and one practical implementation each week.