Provides a persistent, cross-tool memory and team knowledge base with semantic search and real-time analytics.
Configuration
View docs{
"mcpServers": {
"bparpette-mistralhackathonmcp": {
"url": "https://mistralhackathonmcp-ee61017d.alpic.live/",
"headers": {
"QDRANT_URL": "https://your-cluster.qdrant.io:6333",
"SUPABASE_URL": "https://your-project.supabase.co",
"QDRANT_API_KEY": "your_qdrant_api_key",
"MCP_SERVER_PORT": "3000",
"AWS_LAMBDA_FUNCTION_NAME": "atlasmcp-prod",
"SUPABASE_SERVICE_ROLE_KEY": "your_supabase_service_key"
}
}
}
}AtlasMCP acts as a persistent, cross-tool memory for you and your team. It automatically indexes interactions into a shared, semantic memory you can search and refer to from any AI tool, enabling a true collective intelligence across platforms and improving decision making, onboarding, and collaboration.
You connect an MCP client to AtlasMCP to start storing and querying memories. Use the provided MCP tools to add, search, and manage memories, and rely on the shared team memory to access context from any AI tool connected to AtlasMCP. You can search by natural language queries, filter by category or visibility, and review team analytics to understand how information is used and who contributes.
Prerequisites: you need a modern development environment with Node.js for the web frontend and Python for the MCP server. Make sure you have npm and a Python runtime available on your system.
# Local installation steps (example)
# 1) Clone the project
# 2) Install dependencies
uv sync
# 3) Configure environment
cp example.env config.env
# Edit config.env with your API keysThe production MCP server is hosted at a dedicated URL and runs behind an API key for security. You should use the official MCP configuration snippet to connect your clients to the server.
{
"mcpServers": {
"atlasmcp": {
"url": "https://mistralhackathonmcp-ee61017d.alpic.live/",
"headers": {
"Authorization": "Bearer user_d8a7996df3c777e9ac2914ef16d5b501"
}
}
}
}AtlasMCP uses multi-tenant authentication and row-level security to isolate teams and organizations. You control access through permissions at the memory level (private, team, public) and rely on the web app to enforce team boundaries and real-time analytics.
AtlasMCP provides five core tools to manage memories and gain insights across your team.
- add_memory stores intelligent memories with content, category, tags, visibility, and automatic importance detection. It integrates with Qdrant for vector storage and uses multi-tenant authentication via Supabase.
- search_memories performs advanced semantic search with embeddings, supports content filters by category, visibility, and team, and returns relevance and confidence scores while respecting permissions.
- get_team_insights provides real-time analytics such as top categories, most used tags, most active contributors, most accessed memories, and engagement metrics.
- delete_memory securely deletes memories with permission verification and cleans up references automatically.
- list_memories lets you explore the knowledge base with pagination and filters by user and category.
Use AtlasMCP to capture critical information during a project, then query it to surface business impact, technical context, and decisions across your team.
Store intelligent memories with content, category, tags, visibility, and automatic importance detection. Uses Qdrant for vector storage and Supabase for authentication.
Perform semantic searches on memories with content similarity, filters, and confidence scoring while respecting granular permissions.
Provide real-time analytics on top categories, most used tags, active contributors, and engagement metrics.
Securely delete memories with permission verification and automatic reference cleanup.
Explore memory entries with pagination and filters by user and category.