Roadmap · Specialise
AI agents and MCP (TypeScript)
Opens with creditsFor developers who can already call a model API and want to build AI agents that hold up in real use, in Python or TypeScript. By the end you can write the agent loop by hand, pick and use a framework with reason, publish an MCP server, and prove your agent works with evals, guardrails and traces.
Curated by Tauseef Fayyaz10 stages, 71 topics~5 months at a few hours a weekBy the end, you will be able to
- Write a tool-calling agent loop from scratch against one provider API, with typed tools, validated output, clear errors and step, time and cost limits
- Add memory and explicit state to an agent, and choose between chaining, routing, orchestrator-workers and handoffs for a given task, including when not to use several agents
- Build the same agent in LangGraph, the OpenAI Agents SDK or the Claude Agent SDK, and explain what each framework adds over your own loop
- Build, test in the Inspector and publish an MCP server on the current stateless specification, with sensible tools, resources and authorisation
- Ship an agent with an eval suite in CI, prompt injection defences, human approval for risky actions and tracing of every step in production
The path
10 stages, in the order to learn them. Open one to start.
1How agents workUnderstand what makes a program an agent, the loop every agent runs, how agents differ from fixed workflows, and how to decide whether a task needs an agent at all.6 topics · ~2 weeks
- 1What an agent isAn AI agent is a program where a language model decides the next step, runs it through a tool and looks at the result before deciding again.
- 2Workflows versus agentsIn a workflow, you decide the steps in code and the model fills in parts of a fixed path, such as summarise, then classify, then draft a reply.
- 3The agent loopEvery agent runs the same loop.
- 4Ground truth from the environmentAgents get things wrong, so the ones that work check their progress against real results after each action.
- 5When not to build an agentBefore building, run a four-question test.
- 6The parts of an agent systemA working agent system has five parts.
6 in this stageOpen this stage2Tool calling from scratchBuild the agent loop yourself with one provider API and no framework: define tools, read tool calls, return results and errors, handle parallel calls, then read a second provider to see what is universal.7 topics · ~2 weeks
- 1Defining toolsA tool definition has three parts: a name, a description and a JSON Schema for its arguments.
- 2The tool call and result round tripWhen the model wants a tool, it replies with a tool call instead of text: an id, the tool name and the arguments as JSON.
- 3Writing the loop in codeNow write the loop for real with your provider's official SDK in {choice}.
- 4Tool choice and parallel callsTool choice is a request setting that controls whether the model may call tools.
- 5Tool errors the model can recover fromTools fail: an order is not found, an API times out, an argument is invalid.
- 6Descriptions and result sizeMost tool-calling failures are the wrong tool or wrong arguments, and both are usually fixed in the tool description rather than the system prompt.
- 7Reading a second providerOnce your loop works with one provider, read the tool-calling guide of a second one and map the pieces.
7 in this stageOpen this stage- Structured output and validationGet data your code can trust from a model: design JSON Schemas, use strict modes, validate with {choice} tooling, feed errors back for a retry, and enforce business rules in code rather than in the prompt.6 topics · ~2 weeksOpens with credits
- Memory and stateDecide what your agent remembers within a run and across runs: short-term history, context engineering, compaction, explicit typed state, long-term memory stores scoped per user, and checkpoints.7 topics · ~2 weeksOpens with credits
- Planning and multi-agent patternsLearn the reusable shapes for multi-step work, from prompt chains and routers to plan-and-execute, reflection, orchestrator-workers and handoffs, and when splitting work across several agents is worth the cost.7 topics · ~2 weeksOpens with credits
- Agent frameworksRebuild your agent with the frameworks teams actually use in 2026: LangGraph for explicit state and control flow, the OpenAI Agents SDK for agents, handoffs and tracing, and the Claude Agent SDK for agents that work on files and commands. Learn what each adds and when to choose it.7 topics · ~3 weeksOpens with credits
- Build and publish an MCP serverExpose a real system to any MCP host: choose tools, resources and prompts, build a server with the official {choice} SDK on the stateless 2026-07-28 specification, test it in the Inspector, add authorisation for remote use and publish it.8 topics · ~3 weeksOpens with credits
- Evals and testing agentsProve your agent works and keeps working: build an eval set from real tasks, choose code, model and human graders, test tool choice and trajectories with fake tools, measure reliability over repeated trials and gate changes in CI.7 topics · ~2 weeksOpens with credits
- Guardrails, cost and safetyKeep agents safe and affordable: understand prompt injection and the lethal trifecta, give tools least privilege, require human approval for risky actions, sandbox code, enforce step and budget limits, and estimate cost per task.8 topics · ~2 weeksOpens with credits
- Observability and shippingRun your agent in production: trace every step, follow the OpenTelemetry conventions, run long tasks as durable background jobs, make tools idempotent, stream progress, add model fallbacks and turn production failures into new evals.8 topics · ~2 weeksOpens with credits
8 more stages, 58 more topics, free with an account.
You have seen the open stages of the AI agents and MCP roadmap. Every topic it covers is listed above, and an account opens the rest: what each topic means, a question to check yourself, the courses, the practice and Tauseef's advice for every stage. Early access is open: sign up and your account is active shortly after.
Everything here is free to learn from. A few things, like adding a roadmap, keeping a cheat sheet or asking Tauseef a question, use credits. You get 15 when you join, your first roadmap is free, and if you are learning seriously and run out, you can ask for more. Credits only make sure the platform goes to people who are actually learning.
Build a project
A finished project proves more than a certificate.
- Start with this oneintermediate2 semesters · 3 studentsPolicy-Bound Returns and Refunds Agent for Small Online Stores
Tool calling, validation, business rules in code and human approval, measured with simulated customers
- Python
- LLM API with tool calling
- Pydantic
- FastAPI
- PostgreSQL
- intermediate1 semester · 2 studentsDeadline-Tracking Study Agent with Long-Term Memory for University Students
Memory and state done properly: per-user memory, checkpoints, approvals and evals on a real student problem
- Python or TypeScript
- LLM API with tool calling
- LangGraph or the OpenAI Agents SDK
- PostgreSQL with pgvector
- Google Calendar API
- advanced2 semesters · 3 studentsPrompt Injection Red-Team Harness for Tool-Using AI Agents
Evals and security together: attack real agents with injections and measure which defences hold
- Python
- LLM APIs with tool calling
- MCP Python SDK
- Pydantic
- pytest
- advanced2 semesters · 2 studentsSandboxed Spreadsheet Analyst Agent for Small Business Owners
Code execution in a sandbox, cost limits, tracing and trajectory evals on messy real-world spreadsheets
- Python
- pandas
- LLM API with tool calling
- Docker or a hosted code sandbox
- FastAPI
Related on Sefism
Picked from the other libraries by shared topics.
Best practices
See all- Best Practices for Building LLM AppsHow to build LLM apps and agents that hold up in production: prompts, evals, prompt injection, tool permissions, cost, logging and privacy.
- Best Practices for AI Coding AssistantsHow to get reliable code from AI coding assistants and agents: instruction files, task prompts, verification, review, security, dependencies and session hygiene.
- Prompt Engineering Best PracticesHow to write prompts that work the same way on the hundredth input as on the first: clear instructions, structure, examples, reasoning, output control and testing.
Checklists
See all- RAG Production ChecklistA RAG production checklist: chunking, embeddings, hybrid search, reranking, per-user permissions, citations, evals, re-indexing and cost, with how to verify each check.
- LLM Evaluation ChecklistAn LLM evaluation checklist: success criteria, eval datasets, code and model graders, judge calibration, agent evals, safety tests and production monitoring, with how to verify each check.
- AI Feature Launch ChecklistAn AI feature launch checklist: use case fit, evals, UX for errors, disclosure, human override, abuse filters, rate limits, cost caps, prompt privacy, monitoring and rollback.
Project structures
See all- GenAI RAG app (FastAPI) project structureFor developers turning a "chat with your documents" demo into an app people can rely on. A FastAPI backend where every layer has a home: ingestion workers, retrieval, prompts as versioned files, model clients with fallbacks and cost caps, guards, evals that gate CI, and tracing, so a wrong answer can be traced to the step that caused it.
- AI agent with an MCP server project structureFor developers who want to give AI assistants safe access to their own data or API. A TypeScript MCP server built on the official SDK, organised by feature (tasks, projects) with each feature owning its tools, resources and prompts, served over stdio for local hosts and Streamable HTTP with OAuth for remote ones, plus a small agent and the evals that prove a model can use it.
- Python AI agent with LangGraph project structureFor Python developers moving from "call a model in a loop" to an agent that can pause, resume and be debugged. It starts from the official LangGraph template (langgraph.json, a src/agent package, unit and integration tests) and grows into a customer support agent with typed state, one file per node and tool, human approval before refunds, evals that score whole trajectories, tracing, auth and a deployable server image.
Career questions
See all- How do I start learning AI agents as a software engineer?Start with one LLM API and plain code: a loop that calls the model, lets it pick a tool, runs the tool and feeds the result back. Build one small agent that does a real task end to end, add evaluation and guardrails, and only then reach for a framework. Agents are software engineering first.
- Should I learn LangGraph, CrewAI or the OpenAI Agents SDK first, or build AI agents without a framework?Build one small agent without a framework first: the tool-calling loop is about fifty lines and every framework wraps it. Then learn one framework deeply, chosen by what your project needs: LangGraph for explicit state and approvals, the OpenAI Agents SDK for quick assistants with handoffs, the Claude Agent SDK for agents that work on files.
- What is MCP, and do I need to learn it?MCP, the Model Context Protocol, is an open standard for connecting AI applications to tools and data. You do not need to study it in depth, but if you build with LLMs, writing one small MCP server takes an afternoon and teaches you how agents really use tools. Learn the concepts; skip the hype.
Cheat sheets
See all- GenAI app architecture cheat sheetHow a production GenAI app fits together: request flow, layers, RAG, agent loop, memory, caching, guards, evals, tracing, folder layout and launch checklist.
- RAG cheat sheetRetrieval augmented generation end to end: chunking, embeddings, hybrid retrieval, reranking, grounded prompts, evaluation and failure modes.
- AI evals cheat sheetEvals for LLM apps and agents: golden datasets, metrics, LLM-as-judge and its biases, RAG and agent metrics, offline and online evals, CI, human review, tools.
Work with me
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
- 1:1 Mentorship
- Fresh Graduate Mentorship
- Career Guidance
Brand deals, promos and partnerships: tauseeffayyaz@gmail.com