Ein Model Context Protocol-Server, der einem MCP-Client (Claude Desktop, Claude Code oder allem anderen, das MCP spricht) schreibgeschützten Zugriff auf dein Canvas LMS-Konto gewährt.
Er beantwortet zwei Fragen:
„Was belege ich und wie stehe ich?" – aktive Kurse mit aktuellen Noten.
„Was muss ich noch abgeben und wann ist es fällig?" – offene Aufgaben mit Fälligkeitsterminen.
Die Kommunikation nutzt den standardmäßigen stdio-Transport, der Client startet den Server also als Unterprozess. Auf stdout wird außer MCP-Datenverkehr nichts geschrieben.
Node.js 18.17 oder neuer (der Server nutzt das eingebaute fetch)
Ein persönlicher Canvas-Zugriffstoken
Related MCP server: Canvas MCP Server
cd canvas-mcp-server
npm installBeide Variablen sind erforderlich; der Server beendet sich mit einer klaren Meldung, wenn eine fehlt.
Variable | Beschreibung | Beispiel |
| Die Wurzel deiner Canvas-Instanz. Ein abschließendes |
|
| Ein persönlicher Canvas-Zugriffstoken. |
|
Melde dich bei Canvas an.
Gehe zu Konto → Einstellungen.
Klicke unter Zugelassene Integrationen auf + Neuer Zugriffstoken.
Gib einen Zweck und (optional) ein Ablaufdatum an und klicke dann auf Token generieren.
Kopiere den Token sofort – Canvas zeigt ihn nur einmal an.
Der Token trägt deine vollen Canvas-Berechtigungen. Halte ihn aus der Versionsverwaltung heraus und widerrufe ihn auf derselben Einstellungsseite, falls er jemals durchsickert.
Füge den Server zu deiner MCP-Client-Konfiguration hinzu und verweise dabei auf den absoluten Pfad von
src/index.js:
{
"mcpServers": {
"canvas": {
"command": "node",
"args": ["/absolute/path/to/canvas-mcp-server/src/index.js"],
"env": {
"CANVAS_API_URL": "https://asu.instructure.com",
"CANVAS_ACCESS_TOKEN": "your-token-here"
}
}
}
}Claude Desktop – claude_desktop_config.json
(macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\).
Claude Code – claude mcp add canvas --env CANVAS_API_URL=... --env CANVAS_ACCESS_TOKEN=... -- node /absoluter/pfad/zu/canvas-mcp-server/src/index.js
Starte den Client nach dem Bearbeiten der Konfiguration neu.
list_courses_and_gradesJeder Kurs, in dem du als Student aktiv eingeschrieben bist, mit seiner aktuellen Note.
Parameter | Typ | Standard | Beschreibung |
| boolean |
| Auch aktive Einschreibungen aus bereits beendeten Semestern einbeziehen. |
Canvas meldet Noten doppelt, wenn deine Einrichtung Notenperioden verwendet: einmal für die
laufende Periode und einmal für den gesamten Kurs. Das Feld grades.scope gibt an,
welche du gerade betrachtest:
current_grading_period – die Punktzahl deckt die laufende Notenperiode ab, und
course_total_score / course_total_grade tragen die Gesamtkurszahlen.
course_total – die Einrichtung verwendet keine Notenperioden, die Punktzahl ist also
die Kursgesamtpunktzahl.
unavailable – Canvas hat keine Einschreibung mit Notendaten zurückgegeben.
Innerhalb beider Bereiche ignoriert current_* noch nicht bewertete Arbeit, während
final_* unbewertete Arbeit als Null zählt.
{
"courses": [
{
"id": "101",
"name": "Full Stack Web Development",
"course_code": "GIT-411",
"term": "Fall 2026",
"term_start": "2026-08-20T00:00:00Z",
"term_end": "2026-12-18T00:00:00Z",
"enrollment_state": "active",
"grades": {
"current_score": 88.0,
"current_grade": "B+",
"final_score": 80.5,
"final_grade": "B-",
"scope": "current_grading_period",
"grading_period_title": "Fall Term",
"course_total_score": 91.4,
"course_total_grade": "A-"
},
"html_url": "https://asu.instructure.com/courses/101"
}
],
"course_count": 1,
"retrieved_at": "2026-09-01T12:00:00.000Z"
}list_upcoming_assignmentsAufgaben in deinen aktiven Kursen, die noch ausstehen, sortiert nach Fälligkeitsdatum, zuerst die frühesten.
Parameter | Typ | Standard | Beschreibung |
| integer 1–365, oder |
| Wie weit voraus geschaut werden soll. |
| boolean |
| Überfällige, nie abgegebene Arbeit einbeziehen. |
| boolean |
| Ausstehende Arbeit ohne Fälligkeitsdatum einbeziehen. |
| string[] | alle aktiven Kurse | Auf bestimmte Canvas-Kurs-IDs beschränken. |
Eine Aufgabe gilt als ausstehend, wenn sie veröffentlicht, bewertbar und nicht eingereicht, bewertet oder entschuldigt ist. Konkret werden diese herausgefiltert:
alles mit einem Einreichungszeitstempel
Einreichungen im Zustand submitted, pending_review oder graded
entschuldigte Aufgaben
Aufgaben, die bereits eine Punktzahl oder Note tragen (manuelle oder Papier-Eintragung)
not_graded-Aufgaben (Anwesenheitsplatzhalter und Ähnliches)
unveröffentlichte Aufgaben
{
"assignments": [
{
"id": "9004",
"name": "Missed lab writeup",
"course_id": "101",
"course_name": "Full Stack Web Development",
"due_at": "2026-08-28T06:59:00.000Z",
"days_until_due": -4.2,
"overdue": true,
"points_possible": 25,
"submission_types": ["online_upload"],
"submission_state": "unsubmitted",
"missing": true,
"locked": false,
"unlock_at": null,
"lock_at": null,
"html_url": "https://asu.instructure.com/courses/101/assignments/9004"
}
],
"assignment_count": 1,
"courses_checked": 2,
"window": {
"from": "2026-09-01T12:00:00.000Z",
"to": "2026-09-15T12:00:00.000Z",
"include_overdue": true,
"include_undated": false
},
"errors": [],
"retrieved_at": "2026-09-01T12:00:00.000Z"
}Wenn ein Kurs nicht gelesen werden kann – abgeschlossen, eingeschränkt oder anderweitig fehlerhaft –
wird er in errors aufgelistet und die übrigen Kurse liefern weiterhin Ergebnisse.
Paginierung. Canvas paginiert jede Sammlung über den Link-Header. Der
Client folgt rel="next" bei 100 Datensätzen pro Seite, begrenzt auf 20 Seiten pro
Endpunkt, damit eine fehlerhafte Antwort nicht in einer Endlosschleife enden kann.
Nebenläufigkeit. Aufgaben werden von höchstens 5 Kursen gleichzeitig abgerufen, um die Canvas-Ratenlimits nicht zu überschreiten.
Aktuelles Semester. Standardmäßig werden nur Kurse zurückgegeben, deren Semester nicht beendet ist. Das Standard-Semester von Canvas hat kein Enddatum und ist immer enthalten.
Fehler. Canvas-Fehler werden als MCP-Tool-Fehler mit dem Statuscode und der Canvas-eigenen Meldung zurückgegeben, mit einem Hinweis für die häufigen Fälle (401 → falscher Token, 404 → falsche URL).
Schreibgeschützt. Beide Tools sind mit readOnlyHint annotiert. Der Server sendet nur
GET-Anfragen und ändert niemals deine Canvas-Daten.
npm test # 36 tests: API client, grade logic, filtering, and an end-to-end MCP round tripDie Suite verwendet einen fetch-Ersatz mit aufgezeichneten Canvas-Payloads, sodass kein Netzwerk oder
echter Token benötigt wird. Alle Fixture-Daten sind relativ zum Zeitpunkt der Testausführung.
src/
index.js MCP server: tool definitions, schemas, stdio wiring
canvas.js Canvas REST client: auth, pagination, error mapping
courses.js Active-course and grade normalization
assignments.js Outstanding-assignment filtering and due-date windowsMIT
Fetch every course the user is actively enrolled in as a student, together with the current grade for each. When the institution uses grading periods, the score reported is for the grading period in progress and the whole-course total is included alongside it (see the 'scope' field). Defaults to the current term only.
| Name | Required | Description | Default |
|---|---|---|---|
| include_all_terms | No | Include active enrollments from terms that have already ended. |
| Name | Required | Description |
|---|---|---|
| courses | Yes | |
| course_count | Yes | |
| retrieved_at | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true and openWorldHint=true. The description adds useful behavioral nuance beyond those: grading-period behavior, inclusion of the whole-course total alongside the in-progress score, and the 'scope' field for interpretation. This is valuable context for a read-only listing tool.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two concise sentences. The first sentence front-loads the primary purpose, and the second adds an important grading-period nuance. There is no filler or repetition of schema details.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a simple read-only listing tool with one optional parameter, an output schema, and high schema coverage, the description is complete. It explains active enrollment, grading-period behavior, and term defaulting, which is sufficient for an agent to select and invoke the tool correctly.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, and the only parameter, include_all_terms, is well documented in the schema with a clear description and default value. The main description restates the default behavior ('Defaults to the current term only') but adds no additional parameter-level meaning, so the baseline score of 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description states a specific action ('Fetch every course the user is actively enrolled in as a student') and a clear resource ('courses... together with the current grade for each'). It also adds meaningful scope distinctions like 'Defaults to the current term only' and grading-period behavior, which clearly separates it from the sibling assignment-listing tool.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description gives clear context for when to use the tool: when you need active course enrollments and current grades, defaulting to the current term. It does not explicitly mention alternatives or exclusions relative to list_upcoming_assignments, but the course/grade vs. assignment subject matter implies the boundary clearly.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fetch assignments across the user's active courses that are still outstanding — not submitted, not graded, and not excused — with their due dates, sorted soonest first. Past-due work that was never turned in is included by default.
| Name | Required | Description | Default |
|---|---|---|---|
| course_ids | No | Restrict to these Canvas course IDs. Defaults to all active courses. | |
| days_ahead | No | How far ahead to look, in days. Pass null for no upper bound. | |
| include_overdue | No | Include past-due assignments that were never submitted. | |
| include_undated | No | Include outstanding assignments that have no due date. |
| Name | Required | Description |
|---|---|---|
| errors | Yes | Courses that could not be read; the rest of the results are still valid. |
| window | Yes | |
| assignments | Yes | |
| retrieved_at | Yes | |
| courses_checked | Yes | |
| assignment_count | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The annotations already declare readOnlyHint=true and openWorldHint=true, so the description doesn't need to repeat those. It goes beyond the annotations by specifying the filtering logic (not submitted, not graded, not excused), sorting order (soonest first), and default inclusion of overdue items. This adds useful behavioral context without contradicting the annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two sentences, front-loaded with the core purpose, and every clause adds value. It efficiently conveys the tool's scope, filtering criteria, ordering, and default behavior without any fluff.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the output schema exists (as noted in the context), the description doesn't need to explain return values. The tool is relatively simple with 4 optional parameters, and the description plus schema fully cover the behavior. It could be a 5, but the absence of any prerequisites or error conditions keeps it at a solid 4.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The schema already covers 100% of the parameters with descriptions. The tool description adds a bit more by explaining the filtering and sorting behavior, which relates to how the parameters interact (e.g., include_overdue default true). Since schema coverage is high, a baseline 3 is appropriate, but the description's mention of 'sorted soonest first' and 'Past-due work ... included by default' adds context beyond the schema, justifying a 4.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses a specific verb ('Fetch') and identifies the resource ('assignments across the user's active courses') with clear criteria ('not submitted, not graded, not excused'). It also notes the sorting by due date and the inclusion of past-due unsubmitted work, which distinguishes it from the sibling tool (list_courses_and_grades) that focuses on grades and courses.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description makes it clear that the tool is for outstanding assignments, which implies using it when you need to see what's due or overdue. It does not explicitly state when NOT to use it or mention the sibling alternative, but the context is sufficiently clear for most use cases, and the default behavior (including overdue) is disclosed.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.0.0list_courses_and_gradeslist_upcoming_assignmentsScored across 2 tools
The two tools serve clearly distinct purposes: one retrieves course enrollments with grades, the other retrieves upcoming assignments. There is no ambiguity in selecting between them.
Both tools follow a consistent 'list_' prefix followed by a clear noun phrase ('courses_and_grades', 'upcoming_assignments'), making the naming pattern uniform and predictable.
With only two tools, the server feels minimal but not entirely unreasonable for a focused student dashboard scope. The count is at the borderline of being thin, as agents may expect additional related operations.
The tool surface covers the two primary student needs (grades and upcoming assignments) but lacks operations like retrieving individual course details, assignment submissions, or past assignmentscars. This leaves notable gaps for broader academic workflows, though it may suffice for a narrow use case.
Manage your Canvas coursework with quick access to courses, assignments, and grades. Track upcomin…
Read-only access to Epivo's live course catalogue for AI agents.
Search university course materials, your flashcards, quizzes, streak and quota. All tools read-only.
Ask your AI about your Airbnb and VRBO listings, bookings, guest messages and reviews. Read-only.
