dnotitia/akb

Overview

AKB is an organizational memory system for AI agents that scopes documentation, tables, and files within vaults and unifies them via a URI graph. It serves this unified knowledge over the Model Context Protocol (MCP), enabling agents to query structured and unstructured data from a single endpoint.

Best for

Best for
Developers building AI agent systems that need structured, vault-scoped knowledge retrieval

Use cases

• Give an AI agent access to company documentation and policy files scoped to a specific vault
• Provide agents with a unified, graph-linked interface to query tables and documents for data analysis
• Let an agent retrieve files and metadata from multiple vaults through a single MCP connection

How to use

Tools exposed

• akb_list_vaults
• akb_create_vault
• akb_put
• akb_get
• akb_update
• akb_delete
• akb_put_file
• akb_get_file
• akb_delete_file
• akb_create_table
• akb_alter_table
• akb_drop_table
• akb_sql
• akb_browse
• akb_search
• akb_grep
• akb_drill_down
• akb_relations
• akb_link
• akb_unlink

Tested with

Claude Code, Claude Desktop, Cursor, Windsurf, Cline, Continue

Notes

AKB is an organizational memory system for AI agents that scopes documentation, tables, and files within vaults and unifies them via a URI graph. It serves this unified knowledge over the Model Context Protocol (MCP), enabling agents to query structured and unstructured data from a single endpoint.

44 stars on GitHub. Last updated 2026-06-01.

Use cases

• Give an AI agent access to company documentation and policy files scoped to a specific vault
• Provide agents with a unified, graph-linked interface to query tables and documents for data analysis
• Let an agent retrieve files and metadata from multiple vaults through a single MCP connection

Pros

• Unifies docs, tables, and files under a single URI graph for coherent agent access
• Uses vault scoping to keep knowledge boundaries clean and secure
• Exposes knowledge via the standard MCP, making it compatible with many agent frameworks

Cons

• Small community with only 44 stars, indicating early-stage maturity and limited support
• Requires Python runtime and MCP integration, adding deployment overhead
• As an organizational memory tool, it depends on consistent vault structuring to function effectively

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.