About moodle-mcp
moodle-mcp is an MCP server for teachers who use AI assistants to review Moodle assignments and submissions. It retrieves course work, downloads files, and can submit grades or feedback only after teacher confirmation.
Key features
- 01View courses and assignments
- 02Review submission status and content
- 03Download submission files
- 04Save grades and feedback with confirmation
- 05Return submissions for revision with confirmation
- 06Remove HTML and omit participant emails
Use cases
- 01Identify submissions needing teacher review
- 02Read student work across a course
- 03Download files for assignment review
- 04Grade work and return feedback after approval
MCP support
Provides Moodle course, assignment, submission, file, and grading tools to stdio MCP clients through the Moodle web services REST API. Confirmation is required for student-visible grading and submission changes.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y @av-leschinskiy/moodle-mcpMOODLE_URLMoodle site address (wwwroot), e.g. https://moodle.example.orgMOODLE_TOKEN_FILESecretWhere moodle-mcp-login saves the token (path is resolved via homedir())MOODLE_TOKENSecretWeb service token override (CI / smoke tests)MOODLE_TIMEOUTTimeout of a single API request (ms)MOODLE_CHROMEChrome/Chromium binary for moodle-mcp-login (looked up in PATH by default)MOODLE_DEBUG1/true/yes/on enables debug output to stderr and debug level in the host log channelLocal servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
9Saved tool list · updated today
This tool takes no parameters.
Details
From the project READMEMoodle MCP Server
An MCP server for AI assistants (Claude Code, Cursor, Codex and other stdio clients) on top of the Moodle web services REST API. It is built for teachers: the assistant can see courses, assignments and student submissions, download submission files, and save a grade with feedback, acting as the teacher and only after the teacher confirms. Unlike student-oriented Moodle MCP servers, the focus here is reviewing assignments and grading.
The server works with any Moodle site that has the mobile web service
(moodle_mobile_app) enabled: the token is issued the same way as for the Moodle
mobile app, so login through an external OAuth provider with a captcha works too.
Prerequisites
- Node.js ≥ 22.
- Google Chrome or Chromium, for the
moodle-mcp-logincommand. - Mobile app access enabled on the Moodle site (Site administration → Mobile app).
To check:
<site>/admin/tool/mobile/launch.phpshould open rather than return an error.
Quick start
The example below uses https://moodle.example.org; replace it with your Moodle
address exactly as it appears in the browser (scheme, host, path; a trailing
slash does not matter).
-
Log in, once per machine and again whenever a tool returns a login error (Moodle login and server token, in Russian):
MOODLE_URL=https://moodle.example.org npx -y -p @av-leschinskiy/moodle-mcp moodle-mcp-loginA Chrome window opens: log in to the site as usual and solve the captcha yourself. The command prints
Logged in as <name> on <site>and closes the window. Without-p, npx would start the server instead of the login command. Chrome is looked up in PATH; on macOS and Windows, set its path inMOODLE_CHROME. -
Add the server to your client. Claude Code:
claude mcp add moodle -e MOODLE_URL=https://moodle.example.org -- npx -y @av-leschinskiy/moodle-mcpAdd
-s userto enable it for all projects; without it, only for the current one.Codex, with a command or a block in
~/.codex/config.toml:codex mcp add moodle --env MOODLE_URL=https://moodle.example.org -- npx -y @av-leschinskiy/moodle-mcp[mcp_servers.moodle] command = "npx" args = ["-y", "@av-leschinskiy/moodle-mcp"] env = { MOODLE_URL = "https://moodle.example.org" }Cursor and other clients: a stdio server with the command
npx -y @av-leschinskiy/moodle-mcpand theMOODLE_URLvariable:{ "mcpServers": { "moodle": { "command": "npx", "args": ["-y", "@av-leschinskiy/moodle-mcp"], "env": { "MOODLE_URL": "https://moodle.example.org" } } } } -
Check: ask the assistant "show my Moodle courses".
From a repository clone, use node <clone>/dist/index.js and
node <clone>/dist/cli/login.js instead of npx: dist/ is committed, no build
is needed.
What's inside
10 tools in total; 3 of them require user confirmation.
| Category | Tools | Confirmation | Covers |
|---|---|---|---|
| Site | 2 | — | who I am, Moodle version, available functions, my courses |
| Assignments and submissions | 4 | — | course assignments, participants and submission statuses, what needs the teacher across a course, full submission |
| Files | 1 | — | downloading submission files to disk |
| Grading | 3 | 3 | grade and feedback, submission comment, returning a submission for rework |
Full list of names
Site
moodle_list_coursesmoodle_site_info
Assignments and submissions
moodle_get_submissionmoodle_list_assignmentsmoodle_list_pending_submissionsmoodle_list_submissions
Files
moodle_download_file
Grading
moodle_add_submission_comment— confirmationmoodle_revert_submission_to_draft— confirmationmoodle_save_grade— confirmation
The server acts on your behalf: it sees and changes the same things you do in
the Moodle interface. The host asks for confirmation only for writes that students
will see: a grade with feedback, a submission comment, returning a submission to
draft (destructiveHint: true). Downloading a file
also writes, but only to your disk, and needs no confirmation.
Responses are trimmed for the model context: HTML in descriptions and submissions
is converted to plain text, times to ISO, a grade that is not set to null, and
participant emails are never returned.
Configuration
| Variable | Required | Default | Description |
|---|---|---|---|
MOODLE_URL | yes | — | Moodle site address (wwwroot), e.g. https://moodle.example.org |
MOODLE_TOKEN_FILE | — | ~/.config/mcp-moodle/token.json | Where moodle-mcp-login saves the token (path is resolved via homedir()) |
MOODLE_TOKEN | — | — | Web service token override (CI / smoke tests) |
MOODLE_TIMEOUT | — | 30000 | Timeout of a single API request (ms) |
MOODLE_CHROME | — | — | Chrome/Chromium binary for moodle-mcp-login (looked up in PATH by default) |
MOODLE_DEBUG | — | false | 1/true/yes/on enables debug output to stderr and debug level in the host log channel |
The token file has 0600 permissions and stores the site address it was issued
for: if it does not match MOODLE_URL, the server asks you to run
moodle-mcp-login. The server reads the file on every request, so there is no
need to restart the host after logging in again. The token is scrubbed from error
messages, logs and tool responses (including file URLs).
The Chrome profile used for login is kept separate from your main one
(~/.config/mcp-moodle/chrome-profile): site cookies persist, so logging in again
often works without a password.
Development
npm install # also enables the git hooks (.githooks)
npm run dev # tsx, no compilation
npm run build # native TypeScript 7 → dist/
npm run verify # typecheck + lint + tests + docs check
MOODLE_URL=… npm run smoke # read-only tool chain against a live Moodle
Build only with npm run build: devDependencies contain two compiler versions,
details. dist/ is committed; the pre-commit hook builds
and stages it, bumps the patch version and regenerates the tables above.
Architecture and rules are in AGENTS.md (in Russian).
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.