Synaplan MCP
MCP server of the self-hosted Synaplan AI platform: semantic search over your documents, long-term memory and chat
- 903 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The API key opens the platform's knowledge base and long-term memory, which may hold internal documents
- The platform connects mail, calendars and cloud storage, and write actions wait for approval
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/synaplanDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The API key opens the platform's knowledge base and long-term memory, which may hold internal documents
- The platform connects mail, calendars and cloud storage, and write actions wait for approval
This is third-party code. Review the repository files before installing.
What it does
Synaplan is an open platform for assistants built on your documents. One assistant answers on a website, in WhatsApp, by email, in Outlook, Nextcloud and ownCloud with a shared knowledge base and shared rules. Models can be mixed: cloud ones from OpenAI, Anthropic, Google and Mistral beside local ones through Ollama, and every answer shows its model and price. The Synaplan MCP server exposes the platform's knowledge and memory to agents as tools: semantic search over RAG documents, long-term memory and chat. The server works over Streamable HTTP, the hosted address is `https://web.synaplan.com/mcp` and on your own instance it is the `/mcp` path. The API key goes in the `X-API-Key` header and is created in the settings under API Keys. The platform also contains an MCP client and an Anthropic-compatible `/v1/messages` endpoint for coding clients.
Who it is for. Teams that keep a company knowledge base on their own server and want to give Claude Code or Cursor access to it and to the assistant's memory.
Good fit when
- The agent searches the company knowledge base for the refund policy before drafting a customer reply
- Important decisions and preferences are saved to long-term memory and pulled into later sessions
- Data must not leave the building, so the platform is installed on your own server with local Ollama models
- One assistant has to answer the same way on the website, in WhatsApp and in email from a shared base
Not a fit when
- You only need library documentation for an agent: a heavy platform with Docker and 8 GB of memory is overkill
- A server without Docker and without spare memory: the recommended setup will not start
- You need a stable interface: the Synaplan MCP server is marked early access
Example request
Find the returns period section in the Synaplan knowledge base, quote it and remember that EU customers get an answer within a dayMCP
- Transport
- http
- Authentication
- API key
Environment variables
X-API-Keyrequired, secret- Synaplan API key from the settings under API Keys, sent as a header
Limitations
The platform's MCP server and MCP client are marked early access and the interface may change. Running your own instance needs Docker with Compose v2 and 8 GB of RAM, and the first sign-in uses the administrator account from the README, which you should change. The demo without your own key gives 40 messages a day for 14 days. The hosted version opens with a paid plan or an invitation. The stack is Symfony on PHP, a Vue interface, document search through MariaDB VECTOR or Qdrant, and Redis. The license is Apache-2.0.
How to disable. Disconnect the server in the client (`claude mcp remove synaplan`) and delete the key on the API Keys page. Your own instance is stopped with Docker Compose tools.
FAQ
What is the Synaplan MCP server address?
`https://web.synaplan.com/mcp` for the hosted version and the `/mcp` path on your own host for a self-hosted instance. The transport is Streamable HTTP with the key in the `X-API-Key` header.
How do I run Synaplan myself?
Run `git clone https://github.com/metadist/synaplan.git`, enter the folder and run `make up`. http://localhost:5173 opens and the first message starts a demo with no API key.
Can it work with local models?
Yes, the platform connects models through Ollama while cloud ones from OpenAI, Anthropic, Google and Mistral sit alongside. Sensitive tasks can stay on local models.
What does the platform ask before risky actions?
Sending mail and creating tickets wait for your approval and every step is logged. Reads run immediately while write actions pause with an approval card.
How do I connect external MCP servers to Synaplan?
Under Manage, Connections, MCP Servers. The planner pulls live data through `mcp_fetch` and, with write actions enabled for a server, can create tickets and pages through `mcp_action`; destructive tools stay refused.
Discussion
No questions or reviews yet.
Related
MCP serversBrainAPI turns text into a knowledge graph on your server and serves agents over MCP, REST and a console with fact-path answers
A universal MCP CLI client from Apify: persistent sessions, OAuth, JSON output for scripts and shell access to any MCP server for agents
A personal LLM wiki with a knowledge graph, cited RAG and memory, served to Claude Code and Cursor by an MCP server
Self-hosted long-term memory for agents: semantic search, a knowledge graph and autonomous consolidation over MCP and a REST API
Evolution Managed RAG MCP server
Agentic RAG Telegram Bot + MCP Managed RAG
A Cloud.ru teaching MCP server on top of Evolution Managed RAG: hands the agent context from a knowledge base, with a sample GigaChat Telegram bot included
MCP server of the self-hosted EmperorClaw panel: Claude Code and Codex read company knowledge, tasks and customers in scope
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Synaplan (this card) | yes | API key | 903 | |
| MCP Local RAG | no | not required | 430 | local document search with no separate platform, channels, memory or cost tracking |
| MCP Memory Service | no | not required | 2,117 | a standalone memory service for agents without a knowledge base, channels or chat assistant |
README in short
The text opens with a promise of assistants on your documents across channels and five principles: your channels, models, costs, decisions and stack. Next come a quick start with `make up`, capability tables including a row on MCP and Claude Code, and notes on architecture and license.