Scalable AI Management Platform

Synaplan is an open-source AI platform that lets teams connect multiple LLM providers — OpenAI, Claude, Gemini, and local models via Ollama — under a single interface with per-token cost tracking, GDPR-compliant data handling, and self-hosting as a first-class option. The core workflow runs in three layers: a RAG pipeline that ingests PDFs, crawled website URLs, and structured text; a DAG-based routing layer that maps incoming queries to the right model or knowledge source; and a chat widget that embeds on any website without requiring the site owner to write backend code. Audit logs accompany every interaction, satisfying compliance requirements that would otherwise block deployment in regulated European markets.

The memory system is the feature that separates Synaplan from a generic RAG wrapper. Rather than treating each session as stateless, Synaplan captures preferences, tone markers, and factual context as categorized memory nodes, visualized as a neural map that grows with use. The vendor states this enables the model to draft replies that sound consistent with a team’s established voice — useful for support agents or branded chat where session-to-session inconsistency would be noticeable.

Synaplan fits well for European companies that need a widget-based support interface backed by their own document corpus, and for developers who want to experiment with multi-model routing without handing data to a third-party SaaS. CRM and ERP integration for personalized widget answers is a paid-only feature. Where it breaks: the visual DAG routing layer, while accessible, is not a substitute for code when routing logic involves more than a few conditions. Teams that outgrow the canvas typically end up maintaining a parallel scripting layer, which raises the question of why they kept the visual layer at all. MCP server support is described in the v3.6.4 release notes, indicating the platform can act as an MCP host, though the docs describe this as basic at this stage.