AgenticWorks

A community for developers awakening to agentic AI. Hands-on lessons, enterprise-grade context engineering, and a forum that earns its quiet.

Platform

  • Learn
  • Forum
  • Showcase

Project

  • About

Community

  • Network
  • Code of conduct

Field reports

Monthly notes on what shipped, what broke, and what we learned.

© 2026 AgenticWorks. Built in public.

AgenticWorks
LearnShowcaseForumCommunity
Sign in
See it in action

Three rooms. One workshop.

Every tab below is a feature we already shipped or have wired scaffolding for. No vapor. No fluff. Browse before you sign up.

Awakening Q3 2026

Find your inner circle.

Network isn't a directory. It's a matchmaking surface for builders who want operators, mentors, and pod-mates, without the LinkedIn shtick.

  • Profile-matched intros

    Tell us what you're shipping. We'll surface 3 people in this community working on adjacent problems.

  • Mentorship rooms

    Office-hour-style rooms hosted by senior practitioners — sign up for 25 minutes, leave with a real answer.

  • Open-source pods

    Small groups (3-5) collaborating on focused agentic AI projects with clear scope and 4-week cadence.

The awakening

Arise to your potential.

Three rooms in the workshop. Pick where you start — learn, discuss, connect. The community pulls the rest forward.

keep scrolling
Inside the workshop

How agents actually work in production.

The frameworks and disciplines behind agents that go from “works in a notebook” to “works under load.” Each primer goes deep on one; the ladder below shows how they fit together into a production stack.

Google

ADK

Production-grade agents on Google Cloud.

The Agent Development Kit Google ships Vertex agents on: async by default, model-agnostic, with first-class evaluation hooks. The primer walks from quickstart to a deployed multi-tool agent.

Built in public

A curated community building in the open.

Built for people breaking through the talking points and buzzwords and getting to actual collaboration alongside engineers who deploy agents in enterprise settings every week. No fluff, no hype loops, no influencer feed.

Founder note

“The internet has enough tutorials. It needs more workshops, places where you can sit next to someone who's already shipped what you're trying to build.”

AgenticWorks is the workshop I wanted when I first tried to put an agent into production and realized the gap between “works in a notebook” and “works under load with three other agents in the loop” was several disciplines wide. The mission is simple: grow a curated community of engineers who ship agents, then make it easy for everyone else to learn alongside them.

AgenticWorks Founder

What you'll find here

  • Lessons

    Illustrated tracks on ML → agents → production

  • Workshops

Read the primer
LangChain

LangGraph

Stateful, controllable agent loops.

Agents modeled as directed graphs with explicit checkpoints and replay. Stop hoping the LLM does the right thing; design loops you can debug, branch, and replay turn-by-turn.

Read the primer
CrewAI

CrewAI

Role-based multi-agent crews.

Build agents as a team of researcher, writer, and critic, then compose them into workflows that mirror how humans collaborate. The primer walks through a research-and-publish crew.

Read the primer
Microsoft

AutoGen

Multi-agent conversation framework.

Agents talking to agents. Strong for code-generation pipelines, adversarial evaluation, and any flow where two perspectives beat one. We use it as our reference for eval harnesses.

Read the primer

The discipline ladder

Six disciplines that turn a working prompt into a deployable agent. None of them are optional in production.

01Prompt engSTAGE02Context engSTAGE03Harness eng

Monthly live builds + archived recordings

  • Forum

    Threads tagged by stack, 48-hour answer SLA

  • Office hours

    Weekly rooms with senior practitioners

  • Pods

    Small open-source crews on focused 4-week sprints

    • STAGE04Context RAGSTAGE05OrchestrationSTAGE06Eval & obsSTAGE

    • Tighten the input. Constrain the output. Get repeatability before reaching for scale.

    • Manage the window. Summarize, compress, and route; long context is a budget, not a feature.

    • Wrap the model. Tools, structured outputs, retries, and guardrails turn an LLM into something callable from real code.

    • Retrieve the right truth. RAG isn't one pattern: PDFs, tabular data, and tables-on-PDFs each demand a different stack. Match the data shape with embeddings, rerankers, or hybrid.

    • Wire the loop. Planner → worker → critic. State machines beat free-form agent improv every time.

    • Trust by measurement. Golden traces, regression suites, and online dashboards: ship what you can prove.

    1. 01

      Prompt eng

      Tighten the input. Constrain the output. Get repeatability before reaching for scale.

    2. 02

      Context eng

      Manage the window. Summarize, compress, and route; long context is a budget, not a feature.

    3. 03

      Harness eng

      Wrap the model. Tools, structured outputs, retries, and guardrails turn an LLM into something callable from real code.

    4. 04

      Context RAG

      Retrieve the right truth. RAG isn't one pattern: PDFs, tabular data, and tables-on-PDFs each demand a different stack. Match the data shape with embeddings, rerankers, or hybrid.

    5. 05

      Orchestration

      Wire the loop. Planner → worker → critic. State machines beat free-form agent improv every time.

    06

    Eval & obs

    Trust by measurement. Golden traces, regression suites, and online dashboards: ship what you can prove.

    AI/ML community · Built in public

    Welcome to the AI/ML community thatshows you how Agentic Works.

    The workshop where engineers go from buzzwords to working agents alongside teams shipping at enterprise scale.

    Join the community
    See it in action
    AGENTICWORKFLOW
    FoundationsNeurons → LLMsANN · CNNs · Transformers
    Prompts & ContextEngineering the inputFew-shot · Routing · Compression
    Context RAGRetrieval patternsPDF · Tabular · Hybrid
    HarnessesProduction stacksTools · Eval · Guardrails
    ForumTagged threadsQ&A · Show & tell · Patterns
    WorkshopsLive build-alongsStreams · Archive · Drops
    MentorshipOffice hours1:1 · Reviews · Cohorts
    Open sourceBuild in publicRepo · Pods · MIT
    LEARNCOMMUNITY
    Lessons
    10 + growing
    Frameworks
    ADK · LangGraph · CrewAI
    Forum
    Curated for signal
    Cover image for "ADK treats agent development like software engineering. That is the entire trick."

    The ADK deep dive

    ADK treats agent development like software engineering. That is the entire trick.

    Google's Agent Development Kit is the same framework that powers Agentspace and its own production agent surfaces, open-sourced. Here is the complete map: agents, orchestration, tools, MCP, A2A, the dev UI, and a working enterprise build you can deploy with one call.

    YYoha
    Read article
    Cover image for "Portfolio beats certificates."

    Career path

    Portfolio beats certificates.

    Hiring managers don't want another prompt course. They want evidence you can orchestrate rejection loops: eval harnesses, critic gates, and shipped agent workflows in public.

    Keiko Okafor
    Cover image for "MCP breaks when auth and schema versioning are afterthoughts."

    MCP in production

    MCP breaks when auth and schema versioning are afterthoughts.

    Model Context Protocol servers look simple in a tutorial. In enterprise they need auth boundaries, schema versioning, blast-radius limits, and observability. Otherwise they become your next outage.

    Uma Yamamoto
    Cover image for "Better models widen the gap faster than value."

    News analysis

    Better models widen the gap faster than value.

    Every frontier jump improves generation faster than most teams improve verification. Enterprise risk compounds silently: more output, same taste perimeter.

    Freya Wright

    Browse the rest

    Episodic logs are not structured state.

    Memory in production

    Copilot is a distribution channel, not an architecture.

    Conference watch

    Hooks intercept. Skills package judgment.

    Core primitives

    The enterprise agent play is governance, not models.

    Conference watch

    Browse the forum