Agents · LLMs · Knowledge bases · Worldwide
Your own AI agents, your own LLMs and your own knowledge bases — on your own machines, connected worldwide in a decentralised, peer-to-peer network and shared with your whole team. Your files never leave the house.
your agents · your LLMs · your knowledge · worldwide, peer-to-peer
What you get
Most AI tools rent you one of these. MajesticMesh gives your agency all three — under your control, working together in the same chat.
Describe what an agent should do and it's live in seconds — or turn your own software and services into agents. Built once, used by the whole team.
Connect the models you choose — a cloud provider under your contract or a model running in your office — and share them with your team as their own models.
Turn client files, handbooks and project folders into knowledge bases your team can ask — answers with sources, documents stay on your machines.
How it works
A node is the desktop app on a laptop or a small service on a server. It holds your agents, your documents and the connection to your LLM — and your team talks to it directly.
Install the desktop app, or run the node 24/7 on any Linux server. Each account runs up to five nodes for free.
Connect OpenAI, OpenRouter, a local model in Ollama or LM Studio, or anything OpenAI-compatible. Point the node at a folder of documents.
Invite colleagues by email, share a team link, or let everyone with your company domain in. They chat in the browser — no install needed.
Company knowledge
Turn a folder of PDFs, Word files and notes into a knowledge source your team can ask in plain language. The index lives on your node — nothing is uploaded to us or to a vector cloud.
PDF, DOCX, Markdown and text. The node indexes it and keeps watching — new and changed files are picked up on their own.
Every answer names the file and page it came from, so anyone can check it in seconds.
Keep a source to yourself, share it with your team, or hand out only answers — never the documents themselves.
Your own LLMs
We host no models and never see your prompts. You decide which LLMs your agency uses — and who on the team may use them.
OpenAI, OpenRouter, Groq, Mistral or anything OpenAI-compatible — with the account and the terms your company already has.
A local model in Ollama or LM Studio answers without a single prompt leaving your network — ideal for sensitive client work.
Offer a model to your team as its own LLM — text or image generation — without handing out API keys. You decide who may use it and can switch it off any time.
Built for teams
Run a node per office, per client or per department. Your team uses every agent, model and knowledge base on all of them — strangers see nothing.
Every agent, LLM and knowledge base you add starts on Team: you, your nodes and the people you invited. Making anything public is a deliberate switch.
Add members by invitation or by company domain. They sign in with MajesticMesh, and removing someone takes effect on every node.
Turn sharing on or off, set who may chat, send invitations and remove guests — from your account, even for servers without a screen.
A personal link for one person, or a team link for a whole group. Each guest gets their own access you can revoke any time.
One click clones your node onto any Linux box — or docker compose up with a browser admin.
Give each client project its own node with its own agents, models and documents — and invite the client's people as guests.
Chat
There is no chat on this website. Every node serves its own — your browser connects straight to your node. On this direct way, the conversation is encrypted end to end between the two.
Turn on sharing and your node's chat is reachable under its own node.majesticmesh.net address — on your phone, at home, at a client's office.
The browser opens an encrypted connection straight to the node. Only when a network blocks that does our relay pass the traffic along.
Just ask. The chat picks the right agents and knowledge bases, shows its progress, and can create images with an image model.
Worldwide & decentralised
Every node is a full peer. Your offices, servers and laptops find each other across the globe and connect directly — your agents, models and knowledge never sit in someone else's data centre.
A node in Berlin, one in Zurich, a server in New York — they work as one, even behind office firewalls. Your team uses all of them from wherever it is.
Nodes talk to each other directly over their own private network. Our servers only handle sign-in and help nodes find each other — your agents, models and documents are never stored there.
A new office, a new client, a server in another region: start a node and it joins your network in minutes. No migration, no central bottleneck.
Your own AI agents
Your nodes form a private network: every agent you build — and every model and knowledge base behind it — is available to your whole team, wherever it runs.
“An agent that turns any email into formal German.” The builder writes the spec, you refine it in chat, and it's live in seconds.
Add an image-capable model like any other LLM. Pictures are created in the chat and stored on your node.
Any HTTP service becomes an agent with two endpoints: GET /manifest and POST /actions/…. Any language.
Your agents appear as a single MCP server — Claude Desktop, Cursor and Zed see every one of them as a tool.
Every node speaks the OpenAI API — point Cursor, LibreChat or your own software at it and it can use all of your agents.
Change the LLM behind your agents without rebuilding them — from a cloud provider to a model in your own office and back.