2026.10.24
Synaplan: a self-hosted platform for chat, document knowledge, and connected assistants
Synaplan combines multi-model chat, document RAG, an embeddable widget, and MCP in a self-hosted platform. Teams should assess its multi-container deployment, provider keys, data operations, and upgrade process before adoption.
What it is for
Synaplan is an open-source, self-hosted AI platform built with PHP and Symfony. It brings multi-model chat, file-based knowledge, an embeddable website widget, and MCP into one system, while displaying the cost of each answer. It is licensed under Apache-2.0 and is aimed at teams that want control over deployment location, model choice, and data flows.
This is more than a chat page. The README lists providers including OpenAI, Anthropic, Google Gemini, Groq, Mistral, xAI, and HuggingFace, as well as local models through Ollama. Models can be selected per task. The platform also lists media generation, custom instructions, plugins, an OpenAPI REST API, MCP server and client support, and an Anthropic-compatible endpoint.
Who it suits
Synaplan is a reasonable fit for small and mid-sized teams that want to use internal files in AI-assisted work, tool builders, and self-hosting users. Its feature direction is relevant to organizations that already have model-provider accounts, want visibility into usage costs, or need to expose the same assistant through a website and additional channels such as WhatsApp and email.
If the goal is only to obtain a single chat interface quickly, Synaplan’s multi-component structure may be more than necessary. It makes more sense when chat, knowledge retrieval, connectors, and team administration are being planned as one platform rather than when validating one model’s response quality.
Features and deployment considerations
The README provides a Docker Compose deployment path and describes using either provider keys or local Ollama. A lighter Compose configuration omits TTS, Qdrant, and Centrifugo while retaining document search through MariaDB VECTOR. Deployment therefore requires choices about voice features, vector infrastructure, and real-time capabilities.
Before adoption, confirm provider-key handling, file-ingestion boundaries, backup procedures, connector permissions, and exposure to the network. Self-hosting does not by itself complete compliance or security work. The supplied sources do not establish an independent security audit, performance figures, or minimum hardware requirements, so none should be inferred. Local-model users should separately assess model download size, storage, and inference resources.
Version management also deserves attention. The official v5.5.4 release mentions PDF and Office-file reading for web-search results, tasks reading from connected folders, and plugin-platform improvements. Before upgrading, review the release notes, back up the database, uploads, and vector data, and test migrations and connector behavior in a non-production environment. Synaplan’s main value is concentrating multiple AI entry points and knowledge workflows in a deployable platform; its operational complexity should be counted as part of the adoption cost.
Sources
Sources checked: 2026-10-24