One agent. All your company information. You own the data.
Odysseus is one agent over your repos, docs, tickets, and runbooks. It runs on your VPS or on-premises, so you own the data. Nothing trains anyone else's model. By default nothing leaves the box except the model API call you configure. It works with any model's API: you bring the key. From $2,400 USD to install. Care, if you want it, is $450 USD a month.
The codebase cannot leave the building. The agent should not either.
People paste snippets into a chatbot because the answers are useless without our docs.
Policy, client contracts, or plain caution means the repo cannot sit in someone else's product.
We already pay for a model API. We do not want to be locked to one vendor's assistant.
What the agent can see, and what leaves the box.
A static walkthrough. Nothing here is a live index of a client repo.
What it looks like for a US business like yours.
One private agent over the repo, the runbooks, and the tickets.
A company of about 20 people could not send client contracts or the codebase to a vendor chatbot. We installed Odysseus on their VPS, indexed repos, docs, tickets, and runbooks, and connected the model API they already pay for. Answers stay in context. The data stays on their box. Only the configured API call leaves.
Attributed result. Sourced, not a composite.
The box it runs on is the same production VPS pattern, not a new host.
Odysseus does not introduce a NexFlow cloud. It sits on a server you pay for. The Field Note documents a Hetzner CX22 at €4.51/month running n8n with Postgres 16 and Caddy, plus encrypted off-host backups. It also warns that the encryption key has to be set before the first run. Read the Field Note. US companies usually start in a US region on DigitalOcean. The note is the EU path and the hardening rules.
Simple to get going. We do the heavy lifting.
We install it inside your walls, index the sources you name, connect the model key you already have, and leave.
We agree what it may read
Repos, docs, tickets, and runbooks you choose. Client contracts stay in only if you say so.
We install it on your VPS
Or on-premises. You pay for the machine. We do not take a copy.
We connect your model API
OpenAI, Anthropic, or a local model. The key is yours and can be swapped later.
We state what leaves the box
Nothing by default. Only the model API call you configure. Care, if you want it, is the same $450 USD/month retainer.
Everything you get, in plain English.
- One agent over the sources you name
- Install on your VPS or on-premises
- Index of repos, docs, tickets, and runbooks
- Bring-your-own-key for any model API
- A written line on what leaves the box
- No training of someone else's model
- Optional care at $450 USD/month
- Links onward to document AI and CRM automation when you want those outcomes next
Straight-up pricing. No surprises.
Every job starts with a quick 15-minute call so we both know it is a fit. If it is not right for you, we will tell you.
One-off install and index wiring. Optional care is $450 USD/month, the same retainer as the AI assistants lane. There is no separate Odysseus monthly and no fourth price. You pay the model provider and the server provider.
- Install on your VPS or on-premises
- Index wiring for the sources you name
- Any model API, your key
- Data-handling statement in the handover
- Care $450 USD/month, optional
Need the day-to-day assistant too?
AI assistants is the install for an operator who wants OpenClaw on their own machine. Odysseus is for the company codebase and internal docs. Same price spine. Different job.
See full pricingQuestions people ask us.
How much does Odysseus cost?
Install and index wiring start at $2,400 USD one-off. Optional care is $450 USD/month, the same retainer as the AI assistants lane. There is no separate Odysseus monthly.
Where does company data live?
On your VPS or on-premises. Nothing trains anyone else's model. By default nothing leaves the box. The only data that leaves is the model API call you configure, to the provider whose key you brought.
Which model APIs does it work with?
Any model API you bring a key for, including OpenAI, Anthropic, or a local model. You can swap later. The agent is not locked to one vendor's assistant.
Who is this for, versus the AI assistants page?
AI assistants is for an operator who wants a persistent assistant for customers, templates, and day-to-day work. Odysseus is for a technical founder or ops lead at a company of about 5 to 50 people who needs one private agent over repos, docs, tickets, and runbooks. It can hand work onward to document AI and CRM automation. It does not replace them.
One agent. Your walls. Your key.
Tell us what the agent is allowed to read, and which model API you already pay for.