Introduction
Xavier Documentation
Section titled “Xavier Documentation”Xavier is a Rust memory runtime for AI agent workflows.
For local operation, the recommended interface is authenticated HTTP with curl. MCP remains available as an optional transport for IDE-native tooling.
What Xavier Currently Provides
Section titled “What Xavier Currently Provides”- Shared memory storage for agent workflows
- Semantic retrieval through embeddings and indexed memory
- Belief graph support in the runtime and memory modules
- Workspace-aware usage tracking and quotas
- Code indexing and symbol search through
code-graph - Panel UI and optional MCP transport
Current Runtime Shape
Section titled “Current Runtime Shape”The current integrated binary exposes:
- authenticated HTTP routes
- optional MCP transport
- panel UI routes
- code indexing and symbol search
- workspace-aware memory and usage tracking
Current Storage Story
Section titled “Current Storage Story”- Default validated backend:
FileMemoryStore - SurrealDB is present in the codebase and Docker setup
- SurrealDB should currently be treated as optional or future-facing, not as the default validated backend
Current Product Positioning
Section titled “Current Product Positioning”The most accurate public description today is:
- open source memory runtime for agent workflows
- internally validated production candidate
Avoid stronger claims until latency, auth hardening, and monitoring are closed operationally.
Public Data Export
Section titled “Public Data Export”The planned public export pipeline splits Xavier datasets between lightweight GitHub raw manifests and heavier Hugging Face artifacts. See Public Export for the xavier export --public protocol.