Canvas LMS MCP Server (canvas-mcp)
Ask Claude about your Canvas courses, assignments, deadlines, modules, and grades from one place.
接入你的智能体
發布者自己的指令。不是 PlugMCP 端點。
用我們的金鑰與日誌託管一份副本。
uvx canvas-local-mcp==0.1.3Claude Code 沒有本機指令的安裝連結。複製下方指令並在終端機中執行。
CANVAS_BASE_URL=<value> CANVAS_TOKEN=<secret> claude mcp add admin978-canvas-mcp -- uvx canvas-local-mcp==0.1.3在終端機中執行。
{ "mcpServers": { "admin978-canvas-mcp": { "command": "uvx", "args": [ "canvas-local-mcp==0.1.3" ], "env": { "CANVAS_BASE_URL": "<value>", "CANVAS_TOKEN": "<secret>" } } }}claude_desktop_config.json
遠端 URL 還是本機指令? 兩種安裝方式的差異
canvas-mcp is a local-first MCP server for Canvas LMS users (students, instructors, and MCP builders). It turns Canvas REST API actions into MCP tools that work from Claude Code, Claude Desktop, and other MCP-compatible clients.
Status: alpha. Single-user, no warranty, API surface may still shift. File issues if it breaks.
12345.”12345.”syllabus from course 12345.”Prerequisite: Python 3.10+.
In Canvas: Account → Settings → Approved Integrations → + New Access Token. Copy the token shown (Canvas does not show it again later).
From PyPI (recommended):
pip install canvas-local-mcp
Or from source:
git clone https://github.com/admin978/canvas-mcp.git && cd canvas-mcp
python3 -m venv .venv && source .venv/bin/activate
pip install -e .
~/.canvas.envcurl -fsSL https://raw.githubusercontent.com/admin978/canvas-mcp/main/.canvas.env.example -o ~/.canvas.env
chmod 600 ~/.canvas.env
# edit ~/.canvas.env: set CANVAS_BASE_URL (institution root, no /api/v1)
# and paste the token into CANVAS_TOKEN
For token safety guidance (least privilege, file permissions, rotation/revocation, and vulnerability reporting), see SECURITY.md.
Claude Code:
claude mcp add canvas-local -- canvas-local-mcp
Claude Desktop:
~/Library/Application Support/Claude/claude_desktop_config.json%APPDATA%\Claude\claude_desktop_config.json~/.config/Claude/claude_desktop_config.json{
"mcpServers": {
"canvas-local": {
"command": "canvas-local-mcp"
}
}
}
list_courseslist_assignmentslist_moduleslist_announcementsget_pageget_file_infoget_gradesplanner_itemsupcoming_eventstodocanvas-local-mcp-dump downloads course files and related content for offline indexing.
canvas-local-mcp-dump # all active courses
canvas-local-mcp-dump 12345 67890 # specific course IDs
Output goes to ./canvas-dump/ by default. Override with CANVAS_DUMP_DIR=/path/to/dir.
~/.canvas.env (CANVAS_BASE_URL, CANVAS_TOKEN).No demo GIF is currently committed yet. You can still verify the workflow quickly:
After registering the MCP server, ask Claude:
Placeholder: add a short terminal/GIF walkthrough here in a future PR.
Requires Python 3.10+.
pip install -e ".[dev]"
ruff check canvas_local_mcp tests # lint
pytest # tests run against a mocked Canvas API — no token needed
CI runs lint + tests on Python 3.10–3.13 for every push and pull request.
Follow this order to publish a new version (e.g. 0.1.4) consistently across PyPI and the MCP Registry.
Update all three files to the new version string in a single PR:
The CI test test_server_json_version_matches_package and the workflow validation step will fail if any of these three disagree.
mainWait for all CI checks to pass before merging.
canvas-local-mcp to PyPIThe project uses Trusted Publisher (OIDC) for PyPI uploads.
Trigger the PyPI publish workflow (or run python -m build && twine upload dist/* if you have credentials configured) before creating the tag, so the package is available when the MCP Registry fetches it.
vX.Y.Zgit tag v0.1.4
git push origin v0.1.4
This triggers the Publish to MCP Registry workflow, which:
server.json, pyproject.toml, and the tag all declare the same version.server.json to the MCP Registry via GitHub OIDC (no secrets needed).https://pypi.org/project/canvas-local-mcp/https://registry.modelcontextprotocol.io/?q=canvas-mcphttps://github.com/admin978/canvas-mcp/actionsImportant: never reuse or move an existing tag. If you need to re-publish after a mistake, bump to the next patch version (e.g.
0.1.4) and start from step 1.
If this project helps you manage Canvas with Claude, please try the flow above and consider giving it a ⭐ so other students and educators can find it.
MIT — see LICENSE.
Built by AGENTE 404 S.L. · [email protected]
Local-first Canvas LMS MCP server. Stdio transport, ~125 lines of Python.
{ "mcpServers": { "admin978-canvas-mcp": { "command": "uvx", "args": [ "canvas-local-mcp==0.1.3" ], "env": { "CANVAS_BASE_URL": "<value>", "CANVAS_TOKEN": "<secret>" } } }}工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
發布者列出了以下環境變數:CANVAS_BASE_URL, CANVAS_TOKEN。
Canvas 透過 stdio 在本機執行。
上次檢查時間:2026-10-27。 Local-first Canvas LMS MCP server. Stdio transport, ~125 lines of Python.