SYNAPLAN

Own the flow.Open-source AI platform. Self-hosted. Own the flow.

The open AI that answers from your own documents — with the source, a choice of model, and a visible price.PHP 8.5+ / Symfony backend, Vue 3 frontend, MariaDB VECTOR + Qdrant, gRPC inference, OpenAPI contract, MCP server and client. One `make up` away.

[ Try the chat ]Pricing

Apache-2.0 · no licence fee · in the cloud or on your own serversApache-2.0 · Docker · Kubernetes · self-hosted or hosted in Germany

Mission 0427Live demoT+00:00

inbound: outlook/mail
agent: procurement-assistant
knowledge: /contracts /suppliers
model: llama-local
tools: erp.search mail.send
node: de-dus-k8s-03

Approval required

message received

A demo — but every line maps to a real capability.

What you get.

Files stay put
OpenCloud, Nextcloud, Dropbox and ownCloud.online are connected. Nothing has to move.
The price is one click
One click shows the model, the tokens, and what was sent.
You choose the model
The big US models, Mistral from Europe, Ollama on your own server, German GPUs through TrustedTokens. “Again with…” asks the same question of another model.
The same answer
In the chat, on your website, by email, on WhatsApp and Telegram. The code is Apache-2.0 and free of a licence fee, in the cloud or on your own servers.

01 /The same answerIncoming signalsCH-01…06

In the chat, on your website, by email, on WhatsApp and Telegram.WhatsApp, Outlook, mail and website — one AI for every channel.

One question, one knowledge base, one price. The assistant is the same wherever the question arrives.

A GDPR-compliant AI chatbot and the same agents on WhatsApp Business, Outlook, your website widget, desktop and the API. Every message meets the same RAG knowledge base and the same rules — no matter where it came from.

Mail
The same answer by email.
WhatsApp
The same answer on WhatsApp.
Web
The same answer on your website.
API
For your own integrations.
Outlook
An add-in in the mail program: summarise and draft.
Desktop
A program for the desk.

02 /The assistantAgent coreSYS-03

It reads the documents and shows what the answer costs.AI agents with RAG, tools — and a human where it matters.

One click shows the model, the tokens, and what was sent. “Again with…” asks the same question of another model. An email or a change goes out only after a person approves it.

A Synaplan AI agent is not a chat window with a system prompt. It is a pipeline: RAG over your documents, multi-model routing, tools and MCP, a human approval gate and an audit log for every step.

01 understand
Classify intent, language and channel.
02 retrieve
Find the passage in your documents.
03 reason
Local or cloud model — your routing.
04 tool
Calendar, mail, the systems you already have.
05 approval
A person says yes before an email or a change goes out.
06 execute
Do the thing. Log everything.

Knowledge

Answers from your documents, with the source beside them.

Models

US models, Mistral, Ollama on your own server, TrustedTokens.

Tools

For the steps that come before the answer.

Control

Approval before anything goes out. The price stays visible.

api · anthropic-compatible gateway

$ curl https://your-synaplan.example/api/v1/messages \
    -H "Authorization: Bearer $SYNAPLAN_KEY" \
    -d '{"model":"default","messages":[{"role":"user","content":"Summarise /contracts/2026-Q3"}]}'

{ "role": "assistant", "content": "...", "sources": [ "contracts/2026-Q3/frame-agreement.pdf#p4" ] }

03 /Files stay putConnect everythingTOPO-01

The documents stay where they are.A topology, not just a chatbot.

OpenCloud, Nextcloud, Dropbox and ownCloud.online are connected. The AI moves in beside them. Nothing has to move.

The open-source AI platform sits between WhatsApp, Outlook, your website chatbot, Microsoft 365, Nextcloud and MCP. Channels come in, the RAG knowledge base and tools connect around it, every packet is logged.

Files stay put →Connect everything →

Synaplan

The assistant. One answer, with the source and the price.

Hover or tap a node to follow the traffic.

Synaplan

The core. Agents, knowledge, routing, approvals, audit.

Hover or tap a node to follow the traffic.

04 /Self-hostRun it anywhereDEPLOY

In the cloud, or on your own servers.Self-hosted AI: Docker today, Kubernetes tomorrow.

The same software, under Apache-2.0, with no licence fee. If you want to start it yourself, the command is here.

The whole open-source AI platform is Dockerised and Apache-2.0 licensed. Start on a laptop, move to an on-premise server, scale on Kubernetes with the public Helm chart — or let us host it in Germany.

$ curl -fsSL https://raw.githubusercontent.com/metadist/synaplan/main/install.sh | bash

Start Docker Desktop first.
Then paste this in Terminal. Install guide ↗

laptop · docker compose

$ git clone https://github.com/metadist/synaplan.git
$ cd synaplan
$ docker compose up -d

# optional: local models on your own hardware
$ COMPOSE_PROFILES=local-ai docker compose up -d

→ http://localhost:5173

cluster · kubernetes / helm

$ helm install synaplan \
    oci://ghcr.io/metadist/synaplan-charts/synaplan \
    --namespace synaplan --create-namespace \
    -f values.yaml

# GPU inference sidecar (optional)
$ helm install triton oci://ghcr.io/metadist/synaplan-charts/triton

→ deployment/synaplan-web  3/3 ready

01
Laptop
docker compose, ten minutes
02
Server
one Linux box, the deploy/ contract
03
Cluster
Helm chart, Galera, Qdrant
04
Sovereign cloud
openDesk, EU hosting, your keys

Self-host →Deployment docs ↗

05 /Flight logLOG

Missions already flown.

Public institutions, hospitality, care services, industry. Small teams and large ones. Same platform.

LOG-001Operational
Megaherz Yoga
“More time for mindfulness and less effort for organization.”
AI-powered support for course information and bookings.
LOG-002Operational
SANI Pflegedienst
“Caregivers need time for people.”
Automatic answers to frequent questions, reducing team workload.
LOG-003Operational
National Secure Cloud
“Authorities use Synaplan to securely process sensitive data.”
GDPR-compliant protective layer over external AI services.
LOG-004Operational
PlateArt
“Straightforward AI use for support and internal processes.”
Secure AI integration without additional IT burden.

06 /The sourceApache-2.0

Open-source AI platform. Apache-2.0. All of it.

Platform, Helm charts, desktop client, Outlook add-in — public repositories under Apache-2.0. Self-host it, fork it, audit it. Your data stays where you put it.

GitHub stars

868

Forks

License

Apache-2.0

Open repository Read the docs ↗Sovereign edition ↗

engineer mode

$ synaplan inspect

license     Apache-2.0
deployment  docker / kubernetes / managed
backend     PHP 8.5+ · Symfony · FrankenPHP
frontend    Vue 3 · TypeScript · Vite
vectors     MariaDB VECTOR · Qdrant
models      local (ollama) / cloud (6 providers)
rag         enabled
agents      enabled · human approvals
mcp         server + client
api         openapi 3 · anthropic-compatible gateway
channels    mail · whatsapp · web · outlook · desktop · api
lock-in     nope
parrot      operational

07 /Questions

Frequently asked questions

The short answers.

Q-01What exactly is Synaplan?

The open AI that answers from your own documents, with the source, a choice of model, and a visible price. The same answer in the chat, on your website, by email, on WhatsApp and Telegram. In the cloud or on your own servers.

Q-02Is Synaplan a ChatGPT alternative?

Yes. You choose the model, the files stay with you, and the code is Apache-2.0. You are not tied to one provider.

Q-03Is Synaplan GDPR-compliant?

You can run Synaplan in Germany or the EU. With a model on your own server, the question does not leave the building.

Q-04How do I add the chat widget to my website?

A small piece of code on your website. The assistant answers from your pages and documents, in your brand.

Q-05Which AI models does Synaplan support?

The big US models, Mistral from Europe, Ollama on your own server, and German GPUs through TrustedTokens. “Again with…” asks the same question of another model.

Q-06Can I self-host Synaplan?

Yes. The same software, with no licence fee. The start command is on this page, under Self-host.

Q-07What are Synaplan plugins?

Extensions that connect the assistant to the way you already work, for example sorting files or handling documents.

Q-08Which industries and company sizes is Synaplan suited for?

Hospitality, trades, care, industry. Small teams and large ones. The chat widget for the website, the assistant for email and WhatsApp.

07 /Briefing

Frequently asked questions

Everything you need to know about Synaplan — from setup to GDPR compliance.

Q-01What exactly is Synaplan?

Synaplan is an open-source AI platform built on PHP 8.5+. It lets you connect multiple AI models (OpenAI, Anthropic, Groq, Gemini, Ollama), embed a chat widget on any website, search your documents with RAG (retrieval-augmented generation), and extend functionality with business-specific plugins — all in one self-hosted platform.

Q-02Is Synaplan a ChatGPT alternative?

Yes. Unlike ChatGPT, Synaplan runs on your own infrastructure, supports multiple AI models simultaneously, is fully open source, and keeps all data within your control. You are not dependent on a single provider and can switch models at any time.

Q-03Is Synaplan GDPR-compliant?

Absolutely. Synaplan can be self-hosted entirely within Germany or the EU — your data never leaves your servers. With local models via Ollama, no external API calls are made at all. This makes Synaplan one of the few AI platforms that can be operated in full GDPR compliance.

Q-04How do I add the chat widget to my website?

You embed a small JavaScript snippet into your website — just like a Google Analytics or live chat script. The widget connects to your Synaplan instance, uses your custom knowledge base (uploaded documents, FAQs, etc.), and can be styled to match your brand. Setup takes less than 10 minutes.

Q-05Which AI models does Synaplan support?

Synaplan supports OpenAI (GPT-5.5, GPT-5.5 Pro), Anthropic (Claude Opus 4.8), Google Gemini, Groq (ultra-fast inference), and Ollama for local self-hosted models. You can assign different models to different use cases or let the platform route automatically.

Q-06Can I self-host Synaplan?

Yes. Clone the repository, run `make up`, and open http://localhost:5173. The standard install is about 4 GB and uses a cloud model. Local AI (Ollama) is an extra profile and about 5 GB. A local chat model adds about 14 GB.

Q-07What are Synaplan plugins?

Plugins are company-specific AI extensions that connect Synaplan to your existing business processes. Examples include SortX (AI-powered file management), marketing automation, and document workflows. You can use existing plugins or commission a custom plugin tailored to your company.

Q-08Which industries and company sizes is Synaplan suited for?

Synaplan is used by SMBs and enterprises across industries including hospitality, trades & crafts, healthcare, and technology. Website operators use the chat widget for customer service, while larger organizations use the full platform for internal knowledge management, document processing, WhatsApp integration, and email automation.

T-00:00:10

Ready to start?Ready for launch?

Try the chat without an account. The same answer from your own documents, on your website, by email, on WhatsApp and Telegram.

Start the open-source AI platform in two minutes, or self-host it with Docker on yours. Chatbot, agents, RAG — your infrastructure, your rules.

[ Try the chat ]Talk to us

[ Start on web.synaplan.com ]Talk to an engineer