Context Portal

Context Portal (ConPort) is a database-backed MCP server that acts as a memory bank for AI assistants, building a project-specific knowledge graph to enable powerful Retrieval Augmented Generation (RAG) workflows within your IDE.

What It Does

• Stores structured project context (decisions, progress, architecture, tasks, glossaries) in a SQLite database per workspace
• Builds a knowledge graph capturing entities and their relationships
• Provides vector embeddings and semantic search for intelligent context retrieval
• Enables AI assistants to access precise, up-to-date project information for context-aware responses
• Supports prompt caching with compatible LLM providers through structured context
• Manages schema evolution via Alembic migrations
• Replaces file-based memory systems with a queryable, reliable database backend

Key Features

• Multi-workspace support with automatic workspace detection
• Product & Active Context Management for project goals, features, architecture, and current focus
• Decision Logging with full-text search across summaries, rationale, implementation details, and tags
• Progress Tracking with hierarchical task management and item linking
• System Pattern Management for coding patterns and architectural conventions
• Custom Data Storage with category-based organization and FTS search
• Context Linking to explicitly build relationships in the knowledge graph
• History & Versioning for Product and Active Context
• Import/Export to/from Markdown files
• Batch Operations for efficient bulk logging

Available Tools

Product Context

• get_product_context - Retrieve overall project goals, features, and architecture
• update_product_context - Update product context (full or partial patch with __DELETE__ support)

Active Context

• get_active_context - Retrieve current working focus, recent changes, and open issues
• update_active_context - Update active context (full or partial patch)

Decision Logging

• log_decision - Log architectural/implementation decisions with summary, rationale, details, tags
• get_decisions - Retrieve decisions with tag filtering (include_all/include_any)
• search_decisions_fts - Full-text search across decision fields
• delete_decision_by_id - Delete decision by ID

Progress Tracking

• log_progress - Log progress entry/task with status, description, parent_id, item linking
• get_progress - Retrieve progress entries with status/parent filtering
• update_progress - Update existing progress entry
• delete_progress_by_id - Delete progress entry by ID

System Patterns

• log_system_pattern - Log/update system or coding pattern with tags
• get_system_patterns - Retrieve patterns with tag filtering
• delete_system_pattern_by_id - Delete pattern by ID

Custom Data

• log_custom_data - Store/update custom key-value under category (JSON-serializable)
• get_custom_data - Retrieve custom data by category/key
• delete_custom_data - Delete specific custom data entry
• search_project_glossary_fts - FTS search within 'ProjectGlossary' category
• search_custom_data_value_fts - FTS search across all custom data

Knowledge Graph

• link_conport_items - Create relationship link between two items (builds knowledge graph)
• get_linked_items - Retrieve items linked to a specific item with relationship filtering

History & Meta

• get_item_history - Retrieve version history for Product/Active Context
• get_recent_activity_summary - Summary of recent ConPort activity
• get_conport_schema - Retrieve schema of available tools and arguments
• get_workspace_detection_info - Diagnostic info about workspace detection

Import/Export

• export_conport_to_markdown - Export ConPort data to markdown files
• import_markdown_to_conport - Import data from markdown files

Batch

• batch_log_items - Log multiple items of same type in single call

Usage Tips

Automatic Workspace Detection

ConPort auto-detects the correct workspace by searching for project indicators (.git, package.json, pyproject.toml, etc.). Detection is enabled by default and works even if your IDE doesn't expand ${workspaceFolder}. Use --no-auto-detect to disable.

Custom Instructions

The repo includes tailored strategy files for different IDEs:

• roo_code_conport_strategy - For Roo Code
• cline_conport_strategy - For CLine
• cascade_conport_strategy - For Windsurf Cascade
• generic_conport_strategy - Platform-agnostic
• mem4sprint strategy set - Sprint planning patterns with flat categories

Copy the appropriate file's content into your LLM's custom instructions to guide it on ConPort usage.

Initial Setup

Create a projectBrief.md in your workspace root with project overview, goals, features, and architecture. LLM agents will offer to import it as initial Product Context.

Compatible IDEs

Works with any MCP-capable IDE including Roo Code, CLine, Windsurf Cascade, and Cursor.

For detailed architecture, advanced patterns, migration guides, and troubleshooting, see the full documentation in the repository.