Agentโ™ฅ๏ธŽAge
Catalog

Tugra

Official

by dogrucanemek-alt ยท TypeScript

Provenance-aware memory format for AI agents: every claim carries its source, age and boundary.

Model Context Protocol (MCP) Server: com.tugra-ai/tugra

com.tugra-ai/tugra provides an MCP server that serves a provenance-aware memory format for AI agents. It is configured via npx tugra init, which creates a vault, writes a sample fact, and prints a config block to paste into an MCP client. When run in a TTY, npx tugra prints help and exits.

๐Ÿ› ๏ธ Key Features

  • Provenance-aware memory: every claim includes source, age, and boundary
  • Claims stored as files with frontmatter metadata: source, last verification date, shelf life, and boundary/off-limits handling
  • Includes MCP server usage by piping into clients

๐Ÿš€ Use Cases

  • Grounding AI agent responses with claim provenance, freshness, and staleness boundaries
  • Reducing hallucinations by attaching source and verification timing to stored knowledge
  • Working with knowledge-base content in markdown/YAML-frontmatter-based records

โšก Developer Benefits

  • Clear metadata fields for freshness and shelf life via frontmatter
  • TypeScript- and YAML-frontmatter-related workflow alignment indicated by project topics
  • MCP client configuration supported through the printed config block

โš ๏ธ Limitations

  • The provided excerpt does not specify cloud capabilities or full implementation details beyond provenance, aging, and boundary metadata

Topics

agent-memoryai-agentsclaude-desktopcursorfreshnessgroundinghallucination-preventionknowledge-basemarkdownmcpmcp-servermemoryprovenancestalenesstypescriptyaml-frontmatter