DigiMCP (digi)

What's due, attendance, bunk maths, course materials, grades & agentic submissions.

A lightning-fast MCP server and student intelligence CLI for DigiiCampus / CollPoll. One binary, one token, 16 tools — and it plugs into every agent you run.

digi login && digi setup

Every student portal answers the same three questions badly: what's due, how much can I still bunk, and what did I get. You click through nested menus, fight slow pagination, and miss deadlines buried in course tabs.

DigiMCP connects your agent directly to the portal's internal API. Zero browser overhead, cold startup under 6 ms, exact IANA timezone timestamps, in-tree RTK token compaction to keep agent context lean, and 16 native tools for reading records, browsing course materials, and executing validated coursework submissions.

Install & forget

macOS and Linux (Homebrew):

brew install kushvinth/tap/digimcp

(or via cargo)

cargo install digimcp

Linux, macOS or WSL (one-line installer):

curl -fsSL https://raw.githubusercontent.com/kushvinth/DigiMCP/main/scripts/install.sh | sh

Windows (pre-built binary or cargo):

# Pre-built Windows x86_64 binaries available on GitHub Releases:
# https://github.com/kushvinth/DigiMCP/releases/latest

# Or install via cargo:
cargo install digimcp

Then plug it into your agent

# 1. Connect to your college portal (auto-detects session token from your browser)
digi login --portal <yourcollege>.digiicampus.com

# 2. Wire into Claude Code, Codex, Cursor, Gemini CLI, OpenCode
digi setup

Restart your agent. That's it: when it asks "what's due this week?", digi answers in 6 ms with timezone-exact deadlines, attendance with bunk predictions, course slides, grades, or submits feedback forms for you. Nothing to learn, nothing to call, and upgrading digi never changes this.

Command names: The primary CLI binary is digi (fast and intuitive). Binary aliases digii and digimcp are also provided. The crate name on crates.io is digimcp.

Pre-built binaries & release assets · Per-agent manual setup · Dry-run previews

Built-in RTK token compaction (60–80% savings)

Student portals return enormous JSON payloads packed with raw HTML descriptions, redundant navigation fields, inline base64 images, and empty keys. DigiMCP features an in-tree RTK-style token compactor (src/simplify.rs) that runs directly in memory before returning results to your agent:

• HTML stripping: Compresses verbose rich-text descriptions down to concise clean markdown/prose.
• URI neutralization: Replaces multi-kilobyte data URIs with compact placeholders.
• Whitespace normalization: Eliminates wasted runs of indentation and spaces.
• Dead-field pruning: Drops nulls and uninformative keys while preserving lists.

Your agent gets 100% of the relevant data while saving 60% to 80% of context window tokens on every single tool call — with zero ex

... View full README on GitHub