Agent♥︎Age
Catalog

Vestige

Official

by samvallad33 · Rust

Local-first memory for AI agents that reaches backward to find a failure's root cause.

io.github.samvallad33/vestige - MCP Server

Vestige is a local-first memory for AI agents that reaches backward through time to find the quiet changes that caused today's failure. The MCP server exposes model-context capabilities to support long-term memory, failure tracing, and root-cause analysis without cloud reliance. Data remains on-device.

🛠️ Key Features

  • Local-first memory for AI agents with backward tracing
  • Root-cause analysis of failures over time
  • MCP-focused server for model-context orchestration
  • Security: data stays on the user’s machine
  • Rust-based, lightweight footprint (23MB binary)
  • Embeddings, FSRS, and SQLite-backed storage

🚀 Use Cases

  • Debugging AI failure modes by tracing historical changes
  • Maintaining persistent model-context memories for agents
  • Local inference workflows with memory recall
  • Long-term-memory management for cognitive-science research

⚡ Developer Benefits

  • Self-contained, cloud-free deployment
  • Clear MCP-server integration for model-context pipelines
  • Efficient on-device storage and retrieval
  • Open topics relevant to MCP, memory, and tooling

⚠️ Limitations

  • Local-only by design; cloud-based collaboration not provided
  • Readme excerpt implies ongoing readme and docs updates
  • Specific MCP endpoints and schemas not detailed in data provided

Topics

ai-memoryclaudecognitive-sciencecursorembeddingsfsrslong-term-memorymcpmcp-serverneuroscienceonnxrustspaced-repetitionsqlitesveltekitthreejsvscodewebgpulocal-firstsecurity