Challenge course · Claude Code 2.x (protocol: MCP 2026-07-28)

Give your agent tools with MCP (Python)

Build a stateless Model Context Protocol tool server in pure Python — JSON-RPC 2.0 dispatch, schema validation, error isolation, and the 2026-07-28 input_required round-trip.

Advanced5 challenges2–3 hoursComfortable with Python functions and JSONFreeCourse badge: MCP Tool ServerReviewed 2026-09-12English + 中文

Start it with your agent — paste this one sentence:

Read https://flypython.com/skills/flypython/SKILL.md and start the FlyPython course `mcp-server-in-python`.

Needs a coding agent that can run commands and reach the network (chat-only web AIs cannot). Your agent authorizes you with a one-time link, fetches this course's files itself, and you never download anything. Haven't installed the FlyPython Skill? Pick your agent's tool course →

Browse the challenges 中文版 →

What you learnA working MCP tool server implementing the stateless 2026-07-28 specification: no initialize handshake, validated tool arguments, isolated handler failures, and the multi-round input_required flow.

Who it's forComfortable with Python functions and JSON · Python 3.11 or newer available on your PATH

Duration & cost2–3 hours · Free

Challenges

The challenges — read them online

No clone, no account — click a checkpoint to read the full challenge on its own page.

What MCP actually is (2026-07-28)

A stateless JSON-RPC 2.0 surface for tools — no handshake, no sessions, versions in _meta.

Checkpoint l01 · 10 pts · read online →02

The JSON-RPC dispatch contract

Valid requests echo their id; invalid ones get -32600; unknown or removed methods get -32601.

Checkpoint l02 · 10 pts · read online →03

Validation and error isolation

Arguments are untrusted input: validate before execution, and never let a handler crash the transport.

Checkpoint l03 · 10 pts · read online →04

The input_required round-trip

A tool can ask the client a question mid-call — return the requests, finish on the retry carrying answers.

Checkpoint l04 · 10 pts · read online →05

Expose one real tool to your agent

Wrap one function you actually own behind the validated, isolated boundary you just built.

Checkpoint l05 · 10 pts · read online →

Start

How the run works

Your agent drives; the files arrive through the Skill — no download, no clone.

SkillClaude Code 2.x (protocol: MCP 2026-07-28)

The Skill does the fetching

The sentence above starts the course: your agent walks you through the one-time authorization link (compare the code, click Allow), fetches this course's files, and drives the challenges with you. Agents without the Skill installed can work from this brief instead:

Work on the FlyPython challenge "Give your agent tools with MCP (Python)" (course id course-mcp-tools).
Machine-readable brief: https://flypython.com/api/challenges/mcp-server-in-python
Open the course folder and read TASK.md first — it is the contract.
Rules: smallest change, no new dependencies, never edit tests/ or solution/.
Check with python verify.py until its gates pass, then report each claim code to me.

Your agent works under the task contract: smallest change, no new dependencies, never edit solution/. Optional guided mode: COURSE.md.

Files

Files come to you

Course files live at /api/challenges/mcp-server-in-python/files — your agent fetches the manifest and writes each file at its path. There is nothing to download by hand.

How "done" is decided →

Scenario skins

Same skills, a domain you recognize

All three skins run through the same code and the same tests. Pick the one closest to your job before lesson 1.

scenario/01-tools-list.json/requests/01-tools-list.json

Direct tools/list

Listing tools with no prior handshake.

scenario/03-missing-argument.json/requests/03-missing-argument.json

Argument validation

A tools/call missing its required argument.

scenario/04-removed-initialize.json/requests/04-removed-initialize.json

The removed initialize

Calling the method the 2026-07-28 specification removed.

Verification

How “done” is decided

The reviewed mcp-server contract suite: the starter reproduces the listed boundary failures and the solution passes the full suite; wire samples under scenario/requests/ let you probe the protocol by hand.

python verify.py  # from the course folder: checkpoint status + claim codes

Apply this to your own project

Lesson 5 ports the workflow — not the code — to one script you actually own: a three-line task contract, one new failing test made to pass, and a written record of what the tests do not prove. The same loop is described in the AI coding workflow guide and practiced by therunnable examples.

Claim your evidence

When a checkpoint passes, python verify.py prints a deterministic claim code. Your authorized agent submits them for you in one batch; by hand, paste them on yourprogress page — worth 10 points per checkpoint and theMCP Tool Server course badge at five. Claim codes are self-reported evidence, not certificates.

What verification does not prove

Passing the course suite proves the pinned behaviors on the tested inputs — not correctness on tomorrow’s data, not production performance, and not anything about your own project until you write its contract. Repository tests are evidence, not your outcome.

Report teaching drift ↗All courses →