Spectron is a stateless application tier providing persistent, queryable, autonomous memory for AI agents, built directly on SurrealDB. It unifies graph, vector, document, and structured records in one ACID transaction per write. Every fact carries provenance (source, trust, lexical span) and tri-temporal clocks (system time, known time, valid time). Supersession replaces overwrite. Eight pillars and six memory categories (episodic, identity, knowledge, context, instructions, uncertainty) deliver coherence across semantic, lexical, relational, temporal, and spatial dimensions. The MCP server, SDKs (Python, TS, Kotlin, Swift), and harness adapters (LangChain, Claude Code, OpenAI Agents, Vercel AI SDK) enable immediate integration with Cursor, Claude Desktop, Claude Code, and other MCP clients. Currently in invite-only early preview.
Free Trial
How to use Spectron?
To use Spectron, start by joining the waitlist for early preview access. Once granted, download the single Rust binary which includes a built-in MCP server. Connect your MCP client (e.g., Cursor, Claude Desktop, Claude Code) with a single configuration entry. Alternatively, use one of the generated SDKs (Python, TypeScript, Kotlin, Swift) or a harness adapter (LangChain, OpenAI Agents, Vercel AI SDK, n8n, Zapier). Define a Context to scope memory. Ingest conversational turns or documents; extract entities, attributes, and relations automatically via configured LLM. Query using tiered retrieval (direct lookup, response reuse, hybrid retrieval, full-context fallback). Use autonomous mechanisms (reflection, elaboration, consolidation) to deepen understanding between interactions. Monitor the system via the trace graph and built-in commands (spectron entities show, spectron inspect trace).
Spectron 's Core Features
Provenance-first memory with source tracking on every fact
Tri-temporal versioning (system time, known time, valid time) for complete auditability
Supersession instead of overwrite, preserving all historical versions
Six typed memory categories: episodic, identity, knowledge, context, instructions, uncertainty
Eight pillars architecture including reconciliation, elaboration, reflection, consolidation, calibration, and collective memory
