Skip to main content
Glama

📦 Universal Context MCP

The God-Tier Context Preservation Engine for AI Agents

npm version License: MIT

Zero knowledge loss when transitioning between sessions, environments, or agents.

🎯 The Problem

You start a new session with an AI agent (Claude, Gemini, etc.). It spends the first 20 minutes reading files, guessing your architecture, and asking questions you already answered yesterday.

Related MCP server: promem-mcp

⚡ The Solution

Universal Context MCP creates a lossless context bridge. It bundles your project state, uncommitted changes, and architectural intent into a high-density package that "re-animates" any new agent instantly.


✨ Features

  • 🏎️ Zero-Touch Setup: One command auto-configures your global Claude/CLI environment.

  • 🧠 Auto-Bootstrapping: Agents automatically load the previous session's exact focus and blockers.

  • 🕰️ Session Time Machine: A local .session_checkpoints.jsonl ledger tracks the evolution of your project decisions.

  • ⚖️ Precision Token Budgeting: Calculates exact context pressure to prevent agent degradation.

  • ☁️ Cloud Sync: Securely upload context to private GitHub Gists for global cross-machine portability.


📦 Installation

# 1. Install the engine globally
npm install -g universal-context-mcp

# 2. Auto-configure your CLI/Agent environment
context-save --setup

(The setup command automatically injects the MCP server into your global .claude.json or equivalent).


🛠 Usage

1. Manual Export (The "Save State")

Run this in any project root to grab a high-density snapshot and copy it directly to your clipboard:

context-save "Brief description of current focus or blockers"

2. Cloud Export (Cross-Machine)

Moving from your desktop to your laptop? Use a secure GitHub Gist:

context-save "Moving to home laptop" --cloud

3. Agentic Handoff (The "Auto-Bootstrap")

Add this Ironclad Mandate to your project's CLAUDE.md or GEMINI.md:

## Core Mandates
- **Auto-Bootstrap:** At the start of EVERY new session, you MUST immediately call `bootstrap_session_context` to load the previous architectural intent and focus.

When a new agent boots up, it will read this rule, call the MCP tool, and instantly resume your work exactly where you left off.

4. The Autopilot Hook (Zero-Effort Background Sync)

Never forget to save context again. Run this once per project to install a Git hook that silently updates your context in the background every time you commit:

context-save --install-hook

5. Debug Mode

If you need to trace execution speed, path resolution, or token count math, use the debug flag:

context-save --debug

🏗 How It Works Under The Hood

When invoked, the exporter intelligently bundles:

  1. Environment: OS and Path.

  2. Shadow Context: The manual intent or handoff note.

  3. Active Changes: A lightweight git diff --stat.

  4. Structure: A depth-limited, noise-filtered directory tree.

  5. History: The last 5 git log entries.

  6. Core Files: Automatic injection of README.md, CLAUDE.md, package.json, etc.

Available Tools

2 tools
bootstrap_session_contextA

LOAD context: Fetches the latest 'Shadow Context' from the ledger to resume work perfectly.

Inputs: no named parameters.

export_session_contextB

SAVE context: Aggregates project metadata, tree, diffs, and mental model to clipboard and ledger.

Inputs: handoff_message, cloud_sync.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.6/5.0

    Scored across 2 tools

    Disambiguation5/5

    The two tools have clearly distinct purposes: one loads context, the other saves context. There is no functional overlap.

    Naming Consistency5/5

    Both tools follow a consistent verb_noun pattern using snake_case, with 'bootstrap' and 'export' as clear verbs and 'session_context' as the common noun.

    Tool Count3/5

    With only two tools, the server feels minimal. While it covers the core load/save operations, a broader scope might include additional tools for listing or managing contexts.

    Completeness3/5

    The tool set provides basic load and save capabilities, but lacks operations for initializing, deleting, or listing contexts, leaving minor gaps in lifecycle coverage.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      B
      quality
      C
      maintenance
      Enables AI coding agents to maintain persistent project context, including rules, decisions, environment intelligence, and Git history, using a local-first MCP server with automatic project detection and token-efficient retrieval.
      13
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI coding agents to retain and recall project context across sessions through local Markdown storage and hybrid keyword/semantic retrieval, including reusable knowledge cards. It runs offline as a daemon and integrates with MCP-compatible IDEs to continue long-running work without re-explaining decisions or architecture.
      326 npm
      12
      MIT