Self-hosted AI workspace: semantic RAG document search, long-term memory, and AI chat via MCP.
io.github.metadist/synaplan MCP Server
The io.github.metadist/synaplan MCP server is a self-hosted AI workspace that provides semantic RAG document search, long-term memory, and an AI chat interface. It is accessed via MCP, enabling developers to integrate these capabilities into their tooling.
🛠️ Key Features
Semantic RAG document search
Long-term memory
AI chat via MCP
🚀 Use Cases
Searching a document collection using semantic retrieval (RAG)
Persisting and reusing information as long-term memory
Adding AI chat capabilities to an MCP-connected application
⚡ Developer Benefits
Centralizes search, memory, and chat under an MCP server
Supports self-hosted deployment of the workspace capabilities
⚠️ Limitations
Only limited details (name and description) are available; no tool list, topics, or constraints are provided.
Send a message through Synaplan's full AI pipeline — intent classification, web search, RAG over the user's documents, long-term memories, and inference — and return the synthesized answer. This is richer than a raw model call: it uses the authenticated account's knowledge base and memories.
Parameters3
message
string
required
The message or question to send through Synaplan's full AI pipeline (classification, web search, RAG over the user's documents, memories, and inference).
chat_id
integer | null
optional
Optional existing chat id (from list_chats) to continue a conversation. Omit to start a new chat.
agentId
string | integer | null
optional
Optional assistant id or slug (see list_assistants). Pins the reply to that assistant; fails when it is not published, not usable by you, or does not allow connected apps.
Raw schema
{
"type": "object",
"properties": {
"message": {
"type": "string",
"description": "The message or question to send through Synaplan's full AI pipeline (classification, web search, RAG over the user's documents, memories, and inference)."
},
"chat_id": {
"type": [
"integer",
"null"
],
"description": "Optional existing chat id (from list_chats) to continue a conversation. Omit to start a new chat."
},
"agentId": {
"type": [
"string",
"integer",
"null"
],
"description": "Optional assistant id or slug (see list_assistants). Pins the reply to that assistant; fails when it is not published, not usable by you, or does not allow connected apps."
}
},
"required": [
"message"
],
"additionalProperties": false
}
rag_search
Semantic search across the authenticated user's vectorized documents and knowledge base. Returns the most relevant text chunks with similarity scores.
Parameters4
query
string
required
Natural-language search query.
limit
integer
optional
Maximum number of chunks to return (1-50).
min_score
number
optional
Minimum similarity score (0-1) a chunk must reach to be returned.
group_key
string | null
optional
Optional document group key to scope the search to a single document/collection.
Raw schema
{
"type": "object",
"properties": {
"query": {
"type": "string",
"description": "Natural-language search query."
},
"limit": {
"type": "integer",
"minimum": 1,
"maximum": 50,
"default": 10,
"description": "Maximum number of chunks to return (1-50)."
},
"min_score": {
"type": "number",
"minimum": 0,
"maximum": 1,
"default": 0.3,
"description": "Minimum similarity score (0-1) a chunk must reach to be returned."
},
"group_key": {
"type": [
"string",
"null"
],
"description": "Optional document group key to scope the search to a single document/collection."
}
},
"required": [
"query"
],
"additionalProperties": false
}
memory_search
Search the authenticated user's stored long-term memories (preferences, facts, context) using semantic similarity.
Parameters3
query
string
required
Natural-language query describing the memory to look up.
Store a new long-term memory for the authenticated user — a durable fact or preference that Synaplan will recall in future conversations.
Parameters3
category
string
required
Category for the memory, e.g. preferences, work, personal, health.
key
string
required
Short identifier for the memory (min 3 characters), e.g. tech_stack.
value
string
required
The fact or preference to remember.
Raw schema
{
"type": "object",
"properties": {
"category": {
"type": "string",
"description": "Category for the memory, e.g. preferences, work, personal, health."
},
"key": {
"type": "string",
"minLength": 3,
"description": "Short identifier for the memory (min 3 characters), e.g. tech_stack."
},
"value": {
"type": "string",
"minLength": 1,
"description": "The fact or preference to remember."
}
},
"required": [
"category",
"key",
"value"
],
"additionalProperties": false
}
file_ingest
Add a text document to the authenticated user's knowledge base. The content is chunked, embedded, and stored so it becomes retrievable via rag_search.
Parameters3
name
string
required
A human-readable name/title for the document.
content
string
required
The text content to ingest. It is chunked, embedded, and made retrievable via rag_search.
group_key
string | null
optional
Optional group key to scope the document. Pass the same value to rag_search to retrieve only this group; omit it to search the whole corpus.
Raw schema
{
"type": "object",
"properties": {
"name": {
"type": "string",
"description": "A human-readable name/title for the document."
},
"content": {
"type": "string",
"description": "The text content to ingest. It is chunked, embedded, and made retrievable via rag_search."
},
"group_key": {
"type": [
"string",
"null"
],
"description": "Optional group key to scope the document. Pass the same value to rag_search to retrieve only this group; omit it to search the whole corpus."
}
},
"required": [
"name",
"content"
],
"additionalProperties": false
}
rag_similar
Find document chunks similar to a given source message/document id (the `message_id` returned by rag_search).
Parameters2
message_id
integer
required
The source document/message id to find similar chunks for (use a `message_id` returned by rag_search).
limit
integer
optional
Maximum number of similar chunks to return (1-50).
Raw schema
{
"type": "object",
"properties": {
"message_id": {
"type": "integer",
"description": "The source document/message id to find similar chunks for (use a `message_id` returned by rag_search)."
},
"limit": {
"type": "integer",
"minimum": 1,
"maximum": 50,
"default": 10,
"description": "Maximum number of similar chunks to return (1-50)."
}
},
"required": [
"message_id"
],
"additionalProperties": false
}
list_chats
List the authenticated user's conversations (most recently updated first).
Parameters1
limit
integer
optional
Maximum number of chats to return (1-200).
Raw schema
{
"type": "object",
"properties": {
"limit": {
"type": "integer",
"minimum": 1,
"maximum": 200,
"default": 50,
"description": "Maximum number of chats to return (1-200)."
}
},
"additionalProperties": false
}
get_messages
Return the messages of one of the user's chats, in chronological order.
Parameters2
chat_id
integer
required
The id of the chat (from list_chats) to read messages from.
limit
integer
optional
Maximum number of (most recent) messages to return (1-100).
Raw schema
{
"type": "object",
"properties": {
"chat_id": {
"type": "integer",
"description": "The id of the chat (from list_chats) to read messages from."
},
"limit": {
"type": "integer",
"minimum": 1,
"maximum": 100,
"default": 50,
"description": "Maximum number of (most recent) messages to return (1-100)."
}
},
"required": [
"chat_id"
],
"additionalProperties": false
}
list_prompts
List the user's available Synaplan task prompts (internal `tools:*` prompts excluded).
Build AI assistants on your own documents and put them wherever work happens: web, WhatsApp, email, Outlook, Nextcloud and ownCloud. Mix local models with the best cloud AI, see what every answer costs, and decide what the AI is allowed to do – open source, on your server or ours.
Self-hosting is free. The hosted version opens with a paid plan or an invitation code – ask us at team@synaplan.com.
Chat, the cost of every answer, your files, generated media and connected apps – in about 30 seconds.
Why Synaplan
Own your channels. One assistant answers on your website, in WhatsApp, by email, in Outlook and inside Nextcloud and ownCloud – with the same knowledge and the same rules everywhere.
Own your models. Hundreds of models from OpenAI, Anthropic, Google, Mistral, sovereign EU providers or locally via Ollama. Mix them per task: sensitive work stays local, everything else gets the best model for the job.
Own your costs. Every answer shows its model and price. A planner splits bigger requests into steps and spends the expensive model only where it matters.
Own the decisions. Actions like sending mail or creating tickets wait for your approval, and every step is logged.
Own the stack. On-prem, EU cloud or fully air-gapped. No training on your data, no forced telemetry, no feature paywall: self-hosted Synaplan is the same software as our cloud, under Apache-2.0.
Running in production at Megaherz Yoga · SANI Pflegedienst · National Secure Cloud · PlateArt
Quickstart
You need: Docker with Compose v2 and 8 GB of RAM.
bash
git clone https://github.com/metadist/synaplan.git
cd synaplan
make up
Connects to Microsoft 365 · Dropbox · Nextcloud / ownCloud / WebDAV · CalDAV · IMAP · Jira & Confluence · your own MCP servers – how to connect
Under the hood
Stack: Symfony (PHP) backend, Vue frontend, MariaDB VECTOR or Qdrant for document search, Redis, Docker.
Lean core: app, database and Redis. Local AI, office files, web search, voice and more are building blocks you switch on when you need them.
Open by design: OpenAPI REST API, MCP server and client, an Anthropic-compatible /v1/messages endpoint and a plugin system – extend Synaplan without forking it.
Full feature list, every AI provider, all building blocks and developer commands: Platform reference.