DeepMark MCP Server已验证已精选

DeepMark

DeepMark helps teachers deliver rapid, consistent marking with meaningful feedback for every student — in a fraction of the time. What once took a week, now takes one free period.

This MCP server exposes DeepMark's marking engine directly to your assistant, so exam papers can be created, marked and interrogated conversationally — without touching the web app.

What it does

DeepMark marks student answers against a real mark scheme, following the structure an examiner works to. It supports three marking types:

• Deterministic / MCQ — exact-match and multiple choice
• Point-based — discrete credit per mark point
• Levels of response — banded criteria with level placement

Every mark is returned with feedback tied to the specific criteria met and missed, so the reasoning is visible rather than a bare number.

Tools

Papers

• paper_create — build a complete, markable exam paper from pasted text
• paper_confirm — mint a paper from a job awaiting confirmation
• paper_list / paper_get — browse and retrieve papers with question detail
• question_update / markScheme_update — amend questions and mark schemes
• ingestion_get — check the state of a paper-creation or ingestion job

Marking

• submission_mark — mark a set of typed answers against an existing paper
• submission_get — read grading state and per-question marks
• markResult_get — retrieve the marking result for a single answer

Async marking contract

Marking is asynchronous. submission_mark returns a submission_id immediately and grading continues in the background, streaming into the teacher's editor as each question completes.

1. Call submission_mark → receive submission_id
2. Poll submission_get until status is complete
3. Read per-question awarded, max, level and feedback

Allow roughly 15–25 seconds per question. Papers created via paper_create follow the same pattern — poll ingestion_get, then paper_confirm.

Why use it through an assistant

Marking inside a reasoning context means the class set is immediately available to question. Once a set is marked you can ask which mark points the cohort consistently missed, where a levels-of-response band clustered, or which misconception is driving the spread — the diagnostic layer that gets lost when marking is just a number returned.

Setup

Remote server — no local install. Add the URL and authenticate with Google OAuth. Access is scoped to your own account.

Data handling

Uploaded scripts and mark schemes are not used to train models. DeepMark is built in line with the DfE's Generative AI Product Safety Standards.

Built for UK secondary teachers marking GCSE and A-Level extended writing — subject-agnostic and model-agnostic by design.

Website: https://deepmark.co.uk