Your servers
Your cloud, your premises or fully air-gapped hardware. The deployment has no call home, and telemetry is off by default.
Founder2 Bespoke
The sovereign edition of Founder2. The co-founder, the board and your agents deployed on your infrastructure, on your data, with the models you choose. Nothing leaves your perimeter.
The sovereign edition
Founder2 Bespoke is a self-hosted edition of the Founder2 workspace: the living plan, the board, the documents and the co-founder agent, running on infrastructure your company controls. It is built for firms whose data, clients or regulators will not allow a hosted AI.
Your cloud, your premises or fully air-gapped hardware. The deployment has no call home, and telemetry is off by default.
Workspace data, integration keys and model traffic stay inside your network. There is nothing to exfiltrate because nothing is sent back.
Open-weight models like Llama, Qwen, DeepSeek and Mistral on your GPUs, served through vLLM, Ollama or any OpenAI-compatible endpoint — or frontier APIs under your own keys. The co-founder points at the endpoint you give it.
Local AI
When a company asks for a local AI agent, it is asking for three things: the agent runs on infrastructure the company controls, the data it reads never leaves the company, and the models it thinks with are the company's own choice. Founder2 Bespoke is exactly that — a corporate AI agent and assistant, not a chatbot on someone else's cloud.
| What a company gets | Hosted SaaS agents | Building it in-house | Founder2 Bespoke |
|---|---|---|---|
| Where the agent runs | The vendor's cloud | Your infrastructure | Your cloud, premises or air-gap |
| Where your data goes | To the vendor | Nowhere | Nowhere |
| Which models it thinks with | Whatever the vendor ships | Whatever you integrate | Any model you can serve |
| Time to a running workspace | Minutes | A build project of your own | Scoped per deployment |
| Feature updates | The vendor's cadence | Your team writes them | Tested bundles, your approval |
| Human oversight | The vendor's rules | You build the gates | Approval gates, in your logs |
| Compliance paperwork | Shared responsibility | You write it | An audit pack ships with it |
| How it is paid for | Per seat, monthly | Salaries and GPU time | One scoped quote |
What is included
Everything the hosted product does, deployed where you decide. The co-founder, the agents and the data stay on your side of the wall.
The same agent that edits the plan, runs the board and answers from your live data, running inside your perimeter on the model you choose.
Plan, board, roadmap, documents and memory in one workspace your teams share, with version history on every change.
Your other agents join the workspace inside your network, and the co-founder reaches them without anything crossing the boundary.
The same ask-before-it-acts rules as the hosted product. Anything that leaves the workspace waits for a named person's click.
The tools you connect are reached from your network, under keys you hold. They are never registered on our infrastructure.
Wired to your sign-on and access policy during deployment, so the workspace follows the rules you already enforce.
Tested release bundles you apply when you choose, including to sites with no outbound connection at all.
24/7 human support with a named contact, plus a dedicated frontier-powered agent that watches your deployment for bugs and proposes fixes you approve.
Deployment
The same edition, fitted to the perimeter you already run. We confirm the target architecture in the scoping phase.
Into your AWS, Azure or GCP tenancy, or a private cloud you operate. Traffic and storage stay inside your account and your regions.
On your own hardware behind your firewall — Kubernetes, Docker, a VM or bare metal — for firms that keep client data on site. Your team administers it; we support it.
A site with no outbound connection at all. Models run locally and updates arrive as bundles you carry in and apply.
How it works
A deployment is a project we run with your infrastructure team. You see the architecture and the price before anything is installed.
We map your perimeter, your model estate, your identity provider and the teams the workspace will serve. You get a written architecture and a fixed quote before anything is installed.
We deploy with your infrastructure team into your cloud, your premises or an air-gapped site, and connect your identity provider and access policy.
The workspace connects to the tools and stores you allow, and the co-founder points at the models you choose, local or remote, under your keys.
Your team takes over with documentation and training, or we stay on under a maintenance agreement. Updates ship as bundles you apply on your cadence.
Hosted or Bespoke
Founders and small teams start on the hosted plans. Companies that must keep everything in-house run Bespoke. This table is the whole difference.
| Capability | Hosted, Plus and Pro | Bespoke |
|---|---|---|
| Where it runs | Leemer Group cloud, EU residency | Your cloud, your premises or air-gapped |
| Where your data sits | Encrypted on our EU infrastructure | Entirely inside your perimeter |
| Models | Open models via OpenRouter, frontier models on Pro, or your AI subscriptions | Local weights, your inference endpoints, or any API under your keys |
| Agents | Your agents join over MCP, A2A and Poke | The same, plus agents that never leave your network |
| Seats | 1 on Plus, 10 on Pro | Scoped per deployment |
| Updates | Continuous, applied by us | Tested bundles, applied on your cadence |
| Support | Scoped SLA with a named contact | |
| Price | $29 or $99 a month | Custom, scoped per deployment |
Who it is for
SMEs and larger companies alike. A two-partner practice with client files to protect runs the same workspace an enterprise air-gaps. If client data, regulation or policy keeps you out of a shared cloud product, Bespoke is the same workspace without the compromise.
Law firms, fund managers, clinics and auditors whose client files cannot sit on a third party's infrastructure.
A local AI agent for a company this size usually means building it yourself. Bespoke is the same result without the hire.
Divisions that need an AI team under corporate security, procurement and branding rules, not a tool bought on a card.
Product, research and engineering data you would rather never leave the building.
Departments and agencies with residency rules, procurement constraints or air-gapped estates.
Client work delivered with an AI team in the middle, without one client's data ever touching another's.
Firms already running local models or their own inference, that want an agent layer on top of it.
Teams that want frontier-model tooling inside the wall, watched, logged and approved.
Compliance
Every deployment ships with a documentation pack for your compliance team, and the architecture does most of the talking.
Approval gates keep a named person in charge of consequential actions and every action is logged: the human oversight and record-keeping the Act expects from deployers, built in.
When nothing leaves your perimeter there is no processor question to answer. The data map for your records ships with the deployment.
Controls mapped to ISO/IEC 27001 and SOC 2, with the documentation your assessor will ask for.
EU regions, your own racks or no network at all. The residency story is whatever your deployment makes true.
Pricing
There is one edition and one price, and we write both around your deployment. The quote is shaped by seats, deployment targets, your model estate and the support and maintenance you want from us.
What the quote is built from
VPS and server costs passed through at cost, or the workspace set up on hardware you already own. One site, several sites or an air-gapped estate.
Local models billed by the GPU-hour on your estate, or per token on API routes. You pick the route per model and the billing follows it.
Scoping, install, SSO and access wiring, security review and the documentation pack. A fixed number agreed before anything is installed.
Features the product does not have yet, built for your deployment and scoped into the quote. Monthly retention keeps updates and new work landing.
Your branding, your naming and your UI, so the workspace reads as your system, not ours.
24/7 human support with a named contact, plus a dedicated frontier-powered agent that watches for bugs and proposes updates for your approval.
Moving your existing documents, plans and exports into the workspace is part of the deployment, at no charge.
How many people and teams it serves. Sized with you in scoping, never a surprise line later.
Scoped, installed and supported by Leemer Group, Waterford, Ireland.
A self-hosted edition of the full Founder2 workspace: the living plan, the board, the documents and the co-founder agent, deployed on infrastructure your company controls instead of our cloud. It is the same product, inside your perimeter.
An AI agent that runs on infrastructure you control, reads data that never leaves that infrastructure, and thinks on models you choose. 'Sovereign' means your company holds the whole stack: the agent, its memory, its tools and its logs. Founder2 Bespoke is one, for companies rather than individuals.
Companies of every size, from SMEs to large enterprises. A two-partner practice with client files to protect runs the same workspace an enterprise air-gaps. If you simply want the co-founder without infrastructure requirements, the hosted plans from $29 a month are the cheaper way in.
Four things, in one fixed quote: infrastructure (VPS and server costs passed through, or setup on hardware you already own), models (billed by GPU-hour for local weights or per token on API routes, chosen per model), the deployment (install, SSO, security review and documentation) and care (24/7 human support plus the custodian agent). Custom build work and monthly retention are scoped in when you want them.
No. The deployment has no call home: workspace data, integration keys and model traffic stay inside your network. It runs behind an allowlist, and a fully air-gapped install is a supported deployment mode.
Whatever you can serve. Local open-weight models on your own GPUs, an internal inference endpoint your team already runs, or frontier APIs under your own keys. The co-founder treats the model as an endpoint you point it at, and your agents can join over MCP and A2A.
Every deployment ships with an audit-ready pack: a data map of the workspace, controls mapped to ISO/IEC 27001 and SOC 2, GDPR records and EU AI Act deployment notes. Your compliance team gets the documentation, not a promise.
The Act puts duties on the deployer, not only the provider. Approval gates assign human oversight to named people with the authority to hold or refuse an action, every decision is logged on your infrastructure — Article 26 asks for at least six months of retained logs — and the deployment pack covers the transparency, monitoring and record-keeping a high-risk deployment is asked for.
It is that and a step further. Like a private ChatGPT for business, it gives your company an AI assistant on infrastructure you control with no data shared with a vendor. On top of the assistant you get the workspace it works in — the plan, the board, research jobs and other agents — so the AI does work, not only answers.
Same workspace, different perimeter. Hosted Founder2 runs on our cloud from $29 a month. Bespoke runs on your cloud, your premises or air-gapped hardware, is wired to your identity and model estate, and is priced per deployment.
Yes to both. Custom build is scoped into the quote: features the product does not have today, made for your deployment. The workspace carries your branding, your naming and your UI, and a monthly retention keeps new features and updates landing.
We design the deployment with your infrastructure team and do the initial install. Day to day you can administer it yourselves, or keep us on a maintenance agreement: 24/7 human support with a named contact, and a dedicated frontier-powered agent that watches for bugs and proposes updates for your approval.
Migration help is free and part of the deployment. Documents, plans, boards and exports from your current tools are brought in with you before handover, so day one starts from your real data.
Releases ship as tested bundles you apply on your own cadence, including to air-gapped sites. The custodian agent flags what is worth taking; you pick the window and approve what lands.
Custom, scoped per deployment. Seats, deployment targets, your model estate and the support and maintenance you choose shape the quote. Write to support@founderos.app and we will scope it with you.
Tell us about your perimeter, your models and your teams, and we will come back with an architecture and a fixed quote.
Talk to us about Bespoke