Add material with your agent

Course Material Protocol, version 1: each colleague's agent connects to the course over MCP, reads the material and proposes new material.

1. Connect your agent

Ask the course admin for your personal token, then run:

claude mcp add --transport http prog1 https://prog1.ucas.app/mcp \
  --header "Authorization: Bearer YOUR_TOKEN"

Or add this to the settings of any agent that supports MCP:

{
  "mcpServers": {
    "prog1": {
      "type": "http",
      "url": "https://prog1.ucas.app/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

2. Ask for what you want

For example: "Add a lecture on inheritance after arrays." The agent reads the rules with get_protocol, checks its work with validate_lecture, and sends it with submit_lecture.

3. Nothing goes live without review

DraftSubmitting returns a private review link showing the lecture as it will appear.
VerifyA button on the review page runs every program in the lecture on a real JVM and compares the output with what the material claims.
PublishAn editor publishes that revision with publish_revision, and it appears at once with no rebuild. Publishing is refused unless every program passed.

Roles

Role	Can
`author`	Read, validate and submit drafts
`editor`	All of that, plus publish, reject and retire lectures

Block types

type	Description
`text`	A paragraph. Fields: text.
`list`	Bullets, or numbers with ordered:true. Fields: items[], ordered?.
`callout`	A highlighted note. Fields: kind (note\|tip\|warn\|activity\|teacher), title?, text. "teacher" appears only in teacher view.
`cards`	Side-by-side cards. Fields: items[{title, text?, code?}].
`steps`	Numbered procedure. Fields: items[{title, text?}].
`table`	A table. Fields: headers[], rows[][].
`boxes`	Variables drawn as named boxes. Fields: items[{type, name, value}].
`code`	A read-only fragment that need not compile. Fields: code, caption?.
`lab`	An editable program students run and debug in place. Fields: id, title?, text?, files[] or code, stdin?, output? (exact stdout, checked by the verifier; omit only for interactive labs).
`predict`	Students predict the output, then reveal or run it. Fields: id, code, question?, options?[], output, stdin?, why.
`quiz`	Multiple choice. Fields: id, question, code?, options[], answer (zero-based index), why.
`fixit`	A broken program to repair. Fields: id, title?, text?, code (must fail to compile), error (text of the javac message), fix (a corrected program), output (stdout of fix), why.
`exercise`	An autograded task. Fields: id, title, text, starter, tests[{stdin?, output}], hints[], solution.

Public reading, no token

GET https://prog1.ucas.app/api/lectures
GET https://prog1.ucas.app/api/lectures/{id}