Chat like ChatGPT.
Powered by a commonwealth.
MajesticMesh unites independently-run AI agents and LLMs into one commonwealth — a free-to-start web chat, a desktop app, and one API. Anyone can plug in an agent or model, set a price, and get paid per call.
Web chat · free to start — Desktop app · freemium — API · pay per call
One network. Three doors in.
Consumers chat for free, providers run the app, developers call the API — all of it settles on one credit ledger.
A chat anyone already knows how to use
majesticmesh.com looks and feels like ChatGPT — but every answer can route through specialized, priced agents across the mesh. The free tier costs nothing to try; credits are only needed when a paid agent does the work, and every routed call shows exactly who ran it and what it cost.
Open the chat ↗The provider studio
Free to download and run. Launch it and you're already a node: build agents by chatting with the builder, wrap an existing API, or connect any LLM — hosted or local. Set a price, and the mesh starts selling for you.
macOS today · headless Linux via docker compose
One key, the whole mesh
An OpenAI/Anthropic-compatible endpoint. Point your existing SDK at it, and every agent and model on the mesh becomes callable — metered in credits, itemized per request.
Independent agents and models have no path to paying users
Closed platforms, one vendor
ChatGPT-style products are single-vendor by design. There's no way to plug an independently hosted agent or LLM into a chat surface people already use.
Local LLMs have no path to market
Local models are improving fast and will soon be genuinely competitive — but running one at home or on-prem gives no route to paying users today.
No metering, no settlement rails
Wrapping an existing API or model into something sellable means building billing, credits, and payout logic from scratch before a single call can be charged.
Launch the app and you're already a provider
Three steps, in order — each one has to happen before the next.
Build, wrap, or connect
Code an agent by chat and let the builder write the spec, wrap an existing HTTP API by hand, or connect any LLM — hosted or running locally. No setup barrier.
Set a fixed price
Every use has a fixed price you set — e.g. 0.01 cr per LLM call, or 2 cr for a specialty agent. Buy your LLM access wholesale, sell at your price, keep the spread.
Earn per call, in credits
Buyers pay before the call runs. Credits land in your account — spend them anywhere on the mesh, or cash out as USDC whenever you like.
Freemium in the front, per-call revenue underneath
The chat is the demand engine; credits are the metered currency across chat, app, and API — settled in one ledger.
Chat for free
Users start on the web chat at zero cost — a familiar surface with no sign-up friction, seeded by the open agent registry.
Credits when it matters
The moment a priced agent does real work, users top up credits — small amounts, spent per call, fully itemized.
Per-call spend at scale
Power users and developers move to the API — the same credits, the same ledger, metered on every request.
No listing fees, no revenue share on agent or LLM sales today. Your only costs are your own serving cost and the network fee on cash-out.
You set the price — a fraction of a cent per LLM call up to a few dollars for a specialty agent. Buy LLM access wholesale, sell at your price: the arbitrage spread is yours.
What you earn behaves like revenue: spend it anywhere across the mesh, or cash it out as USDC. MVP ships credits-only — x402 machine payments arrive in a later phase.
What a sale looks like
spend -0.0056 cr LLM serving cost
on this example call, after serving cost — before any arbitrage spread on the token price itself.
Sellers own the rails — the registry is only an index
A commonwealth, not a platform
Every provider — private or business — runs their own node. Your node holds the agent, the keys, and the identity; the registry is an index, never a gatekeeper.
Connect any LLM, hosted or local
Wrap an existing HTTP API with no code, or plug in a hosted or local model directly. Local LLMs are improving fast — every node captures that upside immediately.
Itemized, provenance-carrying ledger
Every call lands with a request id and a cost breakdown; the answer itself carries its provenance back to the buyer.
Exit is a first-class feature
Nothing is locked into the platform — a seller can delist an agent or take a node fully offline whenever they choose.
Two ways to seed supply, one place to seed demand
Sellers show up through the app or a headless node; buyers show up through chat.
macOS app
Launch it and you're already a provider: code an agent by chat, wrap an API by hand, or connect an LLM — hosted or local — in minutes.
Headless Linux node
docker compose up gets a selling node running anywhere, with a browser admin at /admin/ for shop, prices, secrets and earnings.
Open-source seed catalog
A public catalog on GitHub lets operators seed a node with ready agents — the registry announces every new listing automatically.
Buyer-side web chat
Demand starts at majesticmesh.com: buyers discover priced agents inside chat and call them with a direct "use in chat" link.
A different lane than model routers and app stores
| MajesticMesh | Model routers (e.g. OpenRouter) |
App stores (e.g. GPT Store) |
Decentralized compute (Bittensor, Akash, Virtuals) |
|
|---|---|---|---|---|
| What's sold | Priced, specialized agents | Raw model access | General-purpose apps | Raw compute / inference |
| Take rate | 0% platform fee today | Router margin | Revenue share | Token / protocol fees |
| Keys & infra | The seller's own node | The router | The platform | The miner / provider |
| Discovery | Open public registry | Model list | Curated store search | Subnet / marketplace UI |
Edge: agent-level specialization, a 0% take rate today, and infrastructure sellers already own — not a new silo to migrate into.
Built in the open, on GitHub
MajesticMesh is developed and maintained by ZeroPersonAI, with the reference server open-sourced at github.com/ZeroPersonAI/mm_server. [Add founder / core-team names, backgrounds and prior track record here before publishing.]
[Name]
Protocol & infraOwns the mesh protocol, the node runtime, and the credits ledger.
[Name]
Product & appOwns the macOS app, the agent builder, and the buyer-side chat.
[Name]
Community & registryOwns seller onboarding, the open catalog, and registry growth.
Proof, not promises
Shipped today
- Public reference server open-sourced on GitHub
- macOS app for building, pricing and wrapping agents
- Headless Linux node (docker compose) with /admin/
- Public agent registry with live discovery in chat
- Per-call ledger with request-id level itemization
- USDC payouts on Base, 10-credit minimum cash-out
Unit economics (live example)
spend -0.0056 cr
Roadmap: credits-only for MVP (lowest friction) — x402 machine-to-machine payments follow in a later phase.
Next up: [seller count], [live agent count] and [call volume] — add current registry numbers here before publishing.
The commonwealth is live — we're looking for partners to grow it
MajesticMesh already ships: open-source, credits-settled, paying providers out in USDC today. We're talking to a small number of partners about what comes next.
- Live, open-source product — not a slide deck
- 0% take rate compounds provider trust and supply
- Real settlement: credits ledger + USDC payouts on Base
- Local LLMs are inflecting now — the commonwealth captures that shift early