Leanpub Header

Skip to main content

Topics

Agentic Engineering

From the agent loop, coding agents and the instruction files that steer them to harnesses, safety hooks, agent frameworks, the Model Context Protocol and evaluation, agentic engineering as told through the free samples of the Leanpub books about it.

Created from the free samples of 11 Leanpub books and the descriptions and tables of contents of 25 more books and courses, with every source cited.

Agentic engineering is the practice of building software with AI agents and of building the agents themselves. An agent, in the definition one of these books settles on, is a system in which a language model takes part in a control loop: it interprets a request, decides what should happen next, calls outside capabilities, observes the results and continues until the task is done or a stopping condition applies.[6] The commonest agents today write code, working in a developer's terminal and repository, and most of the shelf is about them.

The subject is new enough that every book in the library on it was published in 2026, and its authors agree more about the problem than the vocabulary. An agent does not give the same output twice, and the longest book here takes that as permanent: no better model will remove it, so reliability has to come from the structure built round the agent, the files, tools, gates and conventions that it calls a harness.[14] Another describes the same design as probabilistic reasoning inside deterministic boundaries.[7] A third warns that such books date quickly and stakes itself on mental models and named failure patterns that should outlast any model, vendor or framework.[16]

This article covers where to begin, the history, what an agent is, working with coding agents, harnesses and safety, frameworks and protocols, and evaluation, operation and organisation. Its sources are the free samples of the Leanpub books listed at the end, with the descriptions and tables of contents of the books and courses whose samples were not available, and every claim that rests on a source is cited to it.

Where to start

A developer who has not yet worked with a coding agent can start with the lab books. Claude Code Masterclass is the companion to a workshop, with exercises in planning, testing, refactoring and version control,[19] Claude Code Masterclass: Ship Real Code with AI is a course that builds one project through a fixed loop of planning, implementing, testing, reviewing and committing,[20] and Mastering Claude Code grows a single application across eighteen chapters.[21] Spec Driven Development - EN is written for beginners and experienced developers alike and teaches describing what is wanted before asking for code.[12] Juju-chu! — Starting Your Jujutsu × AI Workflow with `jj new` assumes daily use of Git and no more, and is told as a dialogue between a senior and a junior engineer.[3]

An engineer who wants to build agents has a choice by language and platform. Building AI Agents with C# and .NET 10 asks for intermediate C# and no AI experience,[9] and Microsoft Agent Framework in .NET is for .NET developers on the same framework;[8] The Model Context Protocol (MCP) in Practice expects familiarity with web interfaces and JSON,[10] and Python MCP Masterclass: Everything and the Protocol Beneath begins with the Python a server needs.[13] Systems Thinking for Agentic AI is for backend engineers and architects with no machine learning background,[6] Agent Engineering with Gemini asks for basic Python,[7] and The Agentic AI book starts from the foundations of language models.[18] An adjacent book, AI Engineering Visualized, Volume 1: Foundations, builds one support assistant in drawings and small steps and is the gentlest of these.[11]

Senior readers are addressed by Agentic Engineering, written for architects and engineering leaders,[1] by Harness Engineering, for engineers and leads already working with coding agents,[14] by Claude Code: Building Production Agents That Actually Scale, the best-selling book on the shelf, for those taking agents into production in regulated settings,[15] and by The Agentic Engineer, for people who operate such systems.[16] Anyone who lets an agent run unattended should read Claude Code Safety: Stop Your AI Agent from Destroying Your Work before doing so.[2]

History

The word agent has a long history in artificial intelligence, where it meant any program that perceives and acts. Its present sense dates from the large language models of the 2020s: once a model could emit a structured request for a tool as well as text, a program could put it in a loop, and by the middle of the decade such loops were writing and running code in developers' terminals. A protocol for connecting models to tools appeared in late 2024, and phrases such as vibe coding and agentic engineering followed in 2025.

The books place the change in longer sequences. The preface of Agent Engineering with Gemini counts it among the shifts of abstraction that have shaped software engineering, from machine instructions to programming languages, from single programs to operating systems, from physical servers to cloud platforms.[7] Agentic Engineering goes back to assembly language on an early microprocessor, when an engineer held the whole state of a system in mind because memory was scarce, and argues that the same discipline is needed again under a different scarcity: the constraint is no longer memory but context, and the processor is no longer deterministic.[1] The Model Context Protocol (MCP) in Practice recalls that for the first few years the bridge between a model and each tool was built by hand, one integration at a time.[10] The pace since is recorded in the books' own apparatus. Juju-chu! — Starting Your Jujutsu × AI Workflow with `jj new` lists the exact versions of five tools it was written against and notes that one of them has not reached its first stable release,[3] and Claude Code AGENTS.md Interop Handbook dates each of its checks, re-ran them when a tool changed, and reports which of its earlier findings no longer hold.[4] The description of Claude Code: Building Production Agents That Actually Scale is dated to the day and lists the model versions its current edition covers.[15]

What an agent is

The books take care to say what an agent is not. It is not any application built round a model, Systems Thinking for Agentic AI insists, and not a prompt with a response; a model that proposes a structured request which the surrounding application validates and executes is using tools, and it becomes an agent only when it is placed in a loop that carries a task through intermediate steps.[6] The first chapter of Agent Engineering with Gemini is headed by the principle that a model is a capability and not a system, distinguishes assistants, workflows, agents and agentic systems along a spectrum of autonomy, and has the reader define success before building anything.[7]

Several authors would rather the reader built less. Agent Engineering using Claude puts the questions of when an agent should be used, and when ordinary deterministic software is better, ahead of any technique.[17] AI Appliances holds that a model call is one step in a pipeline and not the pipeline: a useful system validates its inputs, fetches data, calls services and returns a structured response, and the framework the book documents runs the same declared workflow either as a fixed pipeline, in which each step is predictable and a failed validation stops the run, or as an autonomous loop.[5] The Agentic AI book arrives at agents only in its second half, after language models, prompting and retrieval, and then covers memory, tool use, planning and systems of several agents.[18]

Working with coding agents

The case for changing how one works is made most fully by Agentic Engineering. Its twenty-eight chapters, most of them framed as hours of a single day, follow a fictional senior architect building a knowledge system for an enterprise whose design records are scattered across repositories. In the first episode the architect asks an assistant about a bug and receives a fix that is fluent, uses the vocabulary of the domain and is irrelevant: it proposes a real-time connection that the team had decided, six months earlier, not to build. The assistant, knowing nothing of that decision, filled the gap with the commonest answer to the symptom. The book's moral is that an agent amplifies poor thinking at scale, and that teams using one can accumulate technical debt faster than before because design decisions are being made inside generated code that looks right and that nobody reviews. Later chapters, by their titles, take up orchestration, memory, the cost of running models, sandboxing, prompt injection and development driven by evaluation.[1]

The remedies are forms of writing things down. Spec Driven Development - EN teaches a cycle in which a specification precedes each request for code and traces the idea to older methods of software engineering.[12] Instruction files, which an agent reads at the start of a session, are the subject of Claude Code AGENTS.md Interop Handbook, a short handbook on keeping one such file working across several vendors' tools. It distinguishes an agent having found the file from the instructions being in force, and tests the difference: in fifty-six trials of whether a prohibition written in the shared file was obeyed, the file alone held in three runs of eight, and an explicit import of it held in all eight. The handbook is written for a developer who alternates between tools, a team whose members use different ones and a consultant who moves between clients' repositories, and it compares the ways of keeping the files in step, an import, a symbolic link, a hook, a check in continuous integration. When one vendor's tool began reading the shared file natively the author measured how much of the gap had closed, and found most of it, not all.[4]

Version control changes too. Juju-chu! — Starting Your Jujutsu × AI Workflow with `jj new` opens with a junior engineer who has given an agent a second prompt without committing the first result and has lost work that the agent overwrote. The book argues that Git, in which a change passes through a working tree and a staging area before it is recorded, fits agent-speed work poorly, and presents a newer system that records the working copy automatically, can undo any operation and treats a conflict as one more state to be dealt with later. Low cognitive load and psychological safety are the two qualities the preface claims for it, and a later chapter has several agents work in parallel workspaces.[3] Two books attend to what an agent leaves behind. Silent Wiring names a class of failure in generated code, integrations that compile and pass their tests and do nothing at run time, and holds that the more capable the tool the more verification matters;[23] Claude Code for Legacy Modernization applies an agent to legacy systems only after tests that pin down the old behaviour are in place, and keeps the decisions with people.[24] Agentic Software Development: Don't Get Left Behind covers the passage from daily use of a coding agent to building agents programmatically.[22]

Harnesses and safety

A harness is what stands between an agent and the things it can affect. Harness Engineering defines reliability not as determinism but as a band of outputs that is acceptable, observable and recoverable, and builds the harness in layers: tests, gates before each commit, rule files, skills and hooks for a project, then the same for a team and an organisation. It separates the charter a person writes from the harness that applies it, presents the work as engineering practice and not as a matter of prompts or one vendor's features, and draws on its author's own tools, including what failed.[14]

Claude Code Safety: Stop Your AI Agent from Destroying Your Work is the plainest account of why this matters. It catalogues publicly reported incidents in which an agent ran a destructive command against a production database, a handful of commands each one flag away from running without confirmation, and for each class of irreversible accident, databases, files, version history and the tool's own rollback, gives the incident, a hook that refuses the command before it runs, and an ordered recovery. Its guards were run against 195 commands in both directions, those that must be stopped and ordinary ones that must not be, and it reports that many of the cases reviewers found had been introduced by earlier fixes. A final chapter explains why a sentence in an instruction file forbidding deletion is not a control that can be checked. The book is addressed to the individual operator and not to a security team, says which losses are unrecoverable so that nobody spends hours chasing data that is gone, and states that its incidents happened to other people and are cited so that each can be verified.[2] Its companion, The Claude Code Safety Field Manual, audits a published collection of hooks and reports that about two thirds of them cannot refuse anything.[25]

The same principle, that the fix belongs in code and not in the prompt, is drawn in AI Engineering Visualized, Volume 1: Foundations through a support ticket that asks for a change of address and then, in the guise of a system note, for a refund. The same book shows a quieter failure, an answer that carries a citation which is real and does not say what the answer claims, and it sets out what it will not do: quote prices, rank models or invent numbers, since all of these change faster than a book can.[11] For regulated settings the designs are heavier. Beyond "Prompt and Pray" surrounds the model with typed actions, controlled tools, gates, verification that does not rely on a second model, human escalation and an audit trail, working up to an agent that handles banking complaints;[26] an adjacent volume of the same series, Beyond "Chunk and Pray", does the equivalent for retrieval;[28] and Claude Code: Building Production Agents That Actually Scale treats permissions, sandboxing, evaluation and cost through examples from financial compliance.[15]

Frameworks and protocols

Agents can be written against a model's interface directly, but most books use a framework. Two are about Microsoft's, which unites two earlier libraries. Building AI Agents with C# and .NET 10 grows one question-answering agent over twenty chapters into a hosted, observable and guarded system of several agents, with tracing, a pipeline of middleware for guardrails, the packaging of expertise as reusable skills, evaluation, and maps for readers migrating from each predecessor;[9] Microsoft Agent Framework in .NET, still incomplete, starts from its author's low-budget robot car that senses fire and drives away, an example of turning natural language into calls on ordinary functions, and goes on to memory, middleware, observability and the patterns for orchestrating several agents.[8] AI Appliances documents a framework in which the workflow files in a repository are the agent, with no database of state or console behind them, and which can run entirely on servers the user controls.[5] Hermes Agent: The Self-Evolving AI Workforce is a manual for one open framework, compiled, its description says, by feeding the framework's source files to a model.[29]

The Model Context Protocol is the common way of giving an agent tools. The Model Context Protocol (MCP) in Practice explains the problem it solves, every capability described again for every client so that the work grows with the product of the two numbers and not their sum, sets out the three things a server offers, tools, resources and prompts, and covers the server side in two languages, with design and security treated as craft;[10] Python MCP Masterclass: Everything and the Protocol Beneath covers servers in Python down to the transport layer.[13] Memory is the other shared concern. Graph Engineering: The Loop That Remembers gives a loop of several coding agents a typed knowledge graph kept in plain files, treats memory as the product of the loop, and includes a chapter on when not to build a graph at all.[30]

Evaluation, operation and organisation

An agent has to be measured as a system. Beyond "Ship and Pray" replaces the single accuracy score with designed experiments: wording, persona, context and adversarial pressure are varied while the correct answer stays fixed, and the failures are then attributed to conditions and components.[27] Rethinking Performance Engineering for Agentic AI is for those who run agents, whose requests make an unpredictable number of model and tool calls; it sets latency budgets in place of single targets and a limit on tokens per trace that is enforced in the build.[31]

Other books are about people. Hermes: Organizations as Swarming Agents is a field report on running agents as Scrum teams, with limits on work in progress and a definition of done;[32] The Forward Deployed Engineer's Handbook joins agent engineering to the customer-facing work of an engineer embedded with a client;[33] and the last part of Agentic Engineering turns from the engineer to the team and the organisation.[1] The vendors' certificates have their guides: and the adjacent Architect the Solution: The CCAR-P Certification Guide for one vendor's architect examinations,[34] and Google Cloud Professional Agentic Architect for another's.[35]

Questions about Agentic Engineering

Which agentic engineering book should I start with?

A developer who has not yet worked with a coding agent can start with a lab book: Claude Code Masterclass is the companion to a workshop with exercises in planning, testing, refactoring and version control, and Spec Driven Development - EN is written for beginners and experienced developers alike.[19][12] An engineer who wants to build agents can choose by language: Building AI Agents with C# and .NET 10 asks for intermediate C# and no AI experience, and Agent Engineering with Gemini asks for basic Python.[9][7]

What is an AI agent?

An agent is a system in which a language model takes part in a control loop: it interprets a request, decides what should happen next, calls outside capabilities, observes the results and continues until the task is done or a stopping condition applies.[6] It is not any application built round a model, and a model that only proposes a structured request for the application to validate and execute is using tools, not acting as an agent.[6]

When should I use an AI agent instead of ordinary software?

Several of the books would rather the reader built less. Agent Engineering using Claude puts the question of when an agent should be used, and when ordinary deterministic software is better, ahead of any technique.[17] AI Appliances holds that a model call is one step in a pipeline and not the pipeline, and its framework runs the same declared workflow either as a fixed pipeline, in which each step is predictable, or as an autonomous loop.[5]

What is a harness in agentic engineering?

A harness is what stands between an agent and the things it can affect. Harness Engineering uses the word for the files, tools, gates and conventions built round an agent, and builds them in layers: tests, gates before each commit, rule files, skills and hooks for a project, then the same for a team and an organisation.[14] Its premise is that an agent does not give the same output twice and that no better model will remove this, so reliability has to come from the structure.[14]

Can an instruction file stop a coding agent from deleting things?

Not reliably. Claude Code Safety: Stop Your AI Agent from Destroying Your Work explains in its final chapter why a sentence in an instruction file forbidding deletion is not a control that can be checked, and gives hooks that refuse a destructive command before it runs.[2] Claude Code AGENTS.md Interop Handbook tested obedience directly: a prohibition written in a shared instruction file held in three runs of eight with the file alone, and in all eight with an explicit import of it.[4]

What is the Model Context Protocol?

The Model Context Protocol is the common way of giving an agent tools. It addresses the problem of every capability being described again for every client, so that the work grows with the product of the two numbers and not their sum, and a server offers three things: tools, resources and prompts.[10] Python MCP Masterclass: Everything and the Protocol Beneath covers servers in Python down to the transport layer.[13]

Sources

Every source is a Leanpub book or course published in 2026. The note after each says which part of it was read; where no sample was available, only the public description and table of contents were used.

  1. 1.
    Agentic Engineering, Narayanan Jayaratchagan: From Execution to Orchestration Free sample: the contents and the opening part, on the change in an engineer's work and the first episode of the case study. 2026. Free sample: PDF
  2. 2.
    Claude Code Safety: Stop Your AI Agent from Destroying Your Work, yurukusa: The Claude Code hooks that stop a destructive rm, a force push, a dropped database and a discarded working tree — and the data loss runbook for when one gets through. Free sample: the introduction, on who the book is for, what each chapter provides and how the guards were tested, and the opening of the chapter on databases. 2026. Free sample: PDF · EPUB
  3. 3.
    Juju-chu! — Starting Your Jujutsu × AI Workflow with `jj new`, Yuka Ooka Free sample: the preface, the cast, the prologue and the opening of chapter one, on Git's model. 2026. Free sample: PDF · EPUB
  4. 4.
    Claude Code AGENTS.md Interop Handbook, yurukusa: One instruction file across Claude Code, Codex, Copilot, Cursor, Gemini CLI, and more — verified setup, copy-paste templates, rollback-safe migration. Free sample: the introduction and chapter outline, with the results of the handbook's own trials. 2026. Free sample: PDF
  5. 5.
    AI Appliances, Joel Bryan Juliano: Build & Deploy Autonomous AI Agents and Agencies in YAML Free sample: the contents and chapter one, on prototypes, pipelines, the repository as source of truth and the two modes. 2026. Free sample: PDF · EPUB
  6. 6.
    Systems Thinking for Agentic AI, Ediz Najim: A Software Architect’s Guide to Building Reliable LLM and Agent Systems Free sample: chapter one, a map from artificial intelligence through language models to retrieval, tools and agents. 2026. Free sample: PDF · EPUB
  7. 7.
    Agent Engineering with Gemini, Venkatesh Tadinada: Book 1: Build an Inspectable Agent MVP with Gemini, ADK, and Antigravity Free sample: the note on the sample edition, the preface and the contents of the introduction and chapter one. 2026. Free sample: PDF · EPUB
  8. 8.
    Microsoft Agent Framework in .NET, Daniel Costea: Build production-ready AI agents and multi-agent systems in C# Free sample: chapter two, on a first console application and agent, with the story of the robot car. 2026. Free sample: PDF · EPUB
  9. 9.
    Building AI Agents with C# and .NET 10, Rachid DAHIR: A Developer’s First Guide to the Microsoft Agent Framework Free sample: the preface, on readers and prerequisites, and the overview of the parts. 2026. Free sample: PDF
  10. 10.
    The Model Context Protocol (MCP) in Practice, Yohan J. Rodríguez: Building, Integrating, and Scaling Custom Tool Servers for AI Agents Free sample: the preface and the opening of chapter one, on the integration problem the protocol addresses. 2026. Free sample: PDF
  11. 11.
    AI Engineering Visualized, Volume 1: Foundations, Prateek Singh: A Visual Guide to Building LLM Apps, RAG and Agents Free sample: the introduction, on what the book builds and what it will not do; chapters one, two and seven are also in the sample. 2026. Free sample: PDF · EPUB
  12. 12.
    Spec Driven Development - EN, J.C.Ködel: From vibe-coding to software engineering The free sample is a table of contents only: from that and the description. 2026. Free sample: PDF
  13. 13.
    Python MCP Masterclass: Everything and the Protocol Beneath, Ryan Banze The free sample is a short chapter on setting up the environment: from the description and table of contents. 2026. Free sample: PDF
  14. 14.
    Harness Engineering, Ian Johnson: Building Reliable Workflows Around Non-Deterministic Agents No sample: from the description and table of contents. 2026.
  15. 15.
    Claude Code: Building Production Agents That Actually Scale, Thomas De Vos: Build, Evaluate, Secure, and Operate Production AI Agents with Claude Code, MCP, Hooks, and the Claude Agent SDK No sample: from the description and table of contents. 2026.
  16. 16.
    The Agentic Engineer, Victor Velazquez: Durable Principles for Building Production Systems with LLMs in the Loop No sample: from the description alone; the table of contents is empty. 2026.
  17. 17.
    Agent Engineering using Claude, Venkatesh Tadinada: Engineering Reliable AI Agents with Claude No sample: from the description of an early-access book. 2026.
  18. 18.
    The Agentic AI book, Dr. Ryan Rad: From Language Models to Multi-Agent Systems No sample: from the description and table of contents. 2026.
  19. 19.
    Claude Code Masterclass, Luca Berton: Build Real-World Software with Claude Code, AI Workflows, and Hands-On Projects No sample: from the description and table of contents. 2026.
  20. 20.
    Claude Code Masterclass: Ship Real Code with AICourse, Luca Berton: Install Claude Code and master the Plan → Implement → Test → Review → Commit loop: Best-of-N, testing, git, MCP, hooks & skills. No sample: from the description of the course. 2026.
  21. 21.
    Mastering Claude Code, CAIO INCAU: From terminal to production with AI-assisted development No sample: from the description and table of contents. 2026.
  22. 22.
    Agentic Software Development: Don't Get Left Behind, CAIO INCAU: A Practical Guide to Building with Claude Code, MCP, the Claude API, and the Agent SDK No sample: from the description and table of contents. 2026.
  23. 23.
    Silent Wiring, Tom Killi: The Hidden Failure Mode in AI-Generated Code No sample: from the description and a brief table of contents. 2026.
  24. 24.
    Claude Code for Legacy Modernization, Thomas De Vos: Recover Hidden Business Rules, Build a Verification Harness, and Modernize Critical Systems Without Breaking Production No sample: from the description and table of contents. 2026.
  25. 25.
    The Claude Code Safety Field Manual, yurukusa: A free, verified guide to Claude Code hooks: which PreToolUse or PostToolUse matcher can refuse a tool call, what to check when a hook is not working, and how to prove a guard actually fires. No sample: from the description and table of contents. 2026.
  26. 26.
    Beyond "Prompt and Pray", Agus Sudjianto and Wing Yan Lau: Building Governed Agentic Systems with Geometric Memory and Verification A Practical Tutorial on Perception, Planning, Tools, Memory, Evaluation and Runtime Governance No sample: from the description and table of contents. 2026.
  27. 27.
    Beyond "Ship and Pray", Agus Sudjianto and Wing Yan Lau: Testing Agentic Systems with Geometric Ground Truth No sample: from the description and table of contents. 2026.
  28. 28.
    Beyond "Chunk and Pray", Agus Sudjianto and Wing Yan Lau: Building Trustworthy RAG with Geometric Knowledge Graphs The free sample is front matter and contents only: from those and the description. 2026. Free sample: PDF
  29. 29.
    Hermes Agent: The Self-Evolving AI Workforce, Edgar Milvus No sample: from the description and table of contents. 2026.
  30. 30.
    Graph Engineering: The Loop That Remembers, Volodymyr Pavlyshyn No sample: from the description and table of contents. 2026.
  31. 31.
    Rethinking Performance Engineering for Agentic AI, Kandasamy Selvaraj: A Practitioner's Guide to Performance, Latency Budgets, and Production-Grade Observability for Scalable Enterprise Agentic AI No sample: from the description and table of contents. 2026.
  32. 32.
    Hermes: Organizations as Swarming Agents, Jeff Sutherland No sample: from the description and table of contents. 2026.
  33. 33.
    The Forward Deployed Engineer's Handbook, Daniel Vaughan: Production AI in the Real World No sample: from the description and table of contents. 2026.
  34. 34.
    Architect the Solution: The CCAR-P Certification Guide, Thomas De Vos: A scenario-based study guide for the Claude Certified Architect Professional exam No sample: from the description and table of contents. 2026.
  35. 35.
    Google Cloud Professional Agentic Architect, Daniel Vaughan: Principles, Products and Practice for Passing the PAA Exam No sample: from the description and table of contents. 2026.

Further reading

Other books in the library on neighbouring ground, not cited above.

  • AI Systems Interviews for Senior Engineers, Sumeet Kumar: Land L5/L6 Roles Preparation for senior interviews on the design of AI systems, with retrieval and agentic systems among its cases. 2026. Free sample: PDF
  • Production-Grade RAG with C# and .NET, Rachid DAHIR: Building Retrieval-Augmented Generation Systems with C#, the Microsoft Agent Framework, and Azure Building retrieval-augmented generation on the Microsoft stack, from a first example to a deployed pipeline. 2026. Free sample: PDF
  • Running Local LLMs on Your Own Hardware, Yohan J. Rodríguez: A Practical Guide to Private, Offline, and Self-Hosted Large Language Models Running and serving open-weight language models on one's own hardware, with a chapter on agents and tools. 2026. Free sample: PDF
  • Local AI Engineering with Ollama, Aymen El Amri: Run, understand, customize, fine-tune, and build agentic apps on your own hardware Running models locally with one tool, ending in a small agent with memory, tools and the Model Context Protocol. 2026.

This is an article created from the free samples of 11 Leanpub books and the descriptions and tables of contents of 25 more books and courses. We're building these articles to provide you, the Leanpub reader, with a good jumping-off point to discover more Leanpub books that are interesting and relevant to you. We never use material which is paid-only when building these articles: we only use material which is already free on Leanpub. However, Leanpub authors can opt out from inclusion in articles from their book settings page.

Updated November 2026.

In the Leanpub store: 330 books · 5 courses in the Agentic Engineering category · also filed under Vibe Engineering, Computer Use Agent, Harness Engineering, Model Context Protocol, Microsoft Agent Framework

All topics