“Legal won’t let us paste client documents into Claude.”
rewired.io → Plan a private chat deployment with the model, hosting, and data path agreed before onboarding.
When legal blocks Claude
Private AI your compliance team can sign off on.
Private AI for your team: chat or code.
Chat endpointchat.rewired.io
API endpointapi.rewired.io
OperatorOperated by Rewired: real people, direct line
What if you controlled what happens behind the cursor?
Make it yours
Preview only
Illustrative conversation
Illustrative preview. Models and hosting confirmed before onboarding, subject to workload fit and availability. Endpoints are issued at onboarding.
Sound familiar?
Your team wants AI. Legal won't sign off on Claude. That's the gap we exist for.
“Our analysts need a chat, not a developer project.”
rewired.io → A private chat interface, with model selection and hosting confirmed before onboarding.
“Our SOC 2 auditor asked where inference runs. We had no answer.”
rewired.io → We confirm the operator, data flow, and hosting location in writing before onboarding.
“We already built on the OpenAI SDK. We can’t rewrite everything.”
rewired.io → Point api.rewired.io at your stack with one env var. LangChain, LiteLLM, and Vercel AI SDK keep working.
01
Agree on the data path
Define which systems handle prompts, who operates them, and what gets logged before onboarding.
02
Chat or code: your team’s covered
Paste documents in chat.rewired.io for legal, ops, and analysts. Connect your app via api.rewired.io with one env var for developers.
03
Start with a scoped pilot
Discuss dedicated capacity as your team grows. Hosting location and infrastructure are confirmed before onboarding, subject to availability.
Built for your buyer
If your industry treats data like a liability, you're ours.
For teams working with sensitive project documents, intellectual property, and commercial information.
Legal & litigation
Client privilege means no pasting into Claude.
Healthcare (future engagements)
Our service is not currently intended for protected health information.
Financial services
Regulators ask where inference runs. You need a real answer.
Client-data agencies
NDAs and SOC 2 audits block shared AI tools.
The data path
Private AI. Defensible infrastructure.
See the proposed path from your team to the model, and what sits outside it.
Your team
Chat or your app
Proposed customer setup
Access gateway
Keys & request policy
Model inference
Agreed hardware & model
Third-party model APIsOutside the proposed inference path
Infrastructure
Owned hardware. Chat or code.
Your team uses chat or connects an app. Behind it, a gateway routes requests to a model running on the agreed hosting infrastructure. We plan that setup around your workload and data-residency needs.
How your private AI runs
Data residency
Location confirmed before onboarding, subject to availability.
Chat or app
You
Paste documents in chat.rewired.io, or one env var for developers.
Same models and privacy boundary whether your team uses the browser or your app calls api.rewired.io.
Onboarding
A clear path from requirements to a pilot.
- 01
Discuss requirements
Choose a document workflow and define what success looks like.
- 02
Confirm the setup
Agree on the model, hosting location, quote, and operating terms.
- 03
Onboard & evaluate
Issue access to the agreed setup and review results with your team.
Access tiers
Pay for data control, not shared cloud risk.
Start with a dedicated pilot arranged for your company, including small teams, subject to workload fit and availability. Developer, Pro, and Team are planned shared-access options; join the waitlist to express interest.
Developer
Planned shared access
1 user
API access for one developer.
From $29/mo
- 1 developer seat (API only)
- Developer API (api.rewired.io)
- Model access confirmed before onboarding
Pro
Planned shared access
1 seat
Individual chat + API access.
From $79/mo
- 1 seat (chat + API)
- Private chat UI (chat.rewired.io)
- API access (api.rewired.io)
Team
Planned shared access
Up to 5 seats
Team chat + API for up to 5 seats.
From $299/mo
- Up to 5 seats (chat + API each)
- Priority queue · multiple API keys
- Model access confirmed before onboarding
Enterprise
Team size by agreement
Dedicated pilot for your company
By quote
- Small teams welcome
- Dedicated hosting subject to availability
- One-time setup fee + monthly minimum
- Larger models assessed for your workload
- Fine-tune hosting · DPA on request
Shared access is planned; the prices above are starting prices, and joining the waitlist does not reserve capacity or guarantee a launch date. Final price, included usage, rate limits, support, and hosting location are agreed before onboarding based on workload, scope, and availability. Dedicated pilots are quoted per customer, with a one-time setup fee and monthly minimum confirmed in your proposal.
Security review
The questions your auditor asks first.
Straight answers, including where we fall short. If a question you need answered is missing, ask it in the form below and we will answer it in writing.
Can we start with a small team?
Yes. Enterprise is a dedicated pilot for your company, including small teams, with no six-seat minimum. Each pilot is arranged per customer, subject to workload fit, hosting location, and availability. The proposal confirms a one-time setup fee and monthly minimum. Developer, Pro, and Team are planned shared-access options; the waitlist registers interest and does not reserve capacity or guarantee a launch date.
Where does inference physically run?
Tell us your data-residency and compliance requirements. We’ll confirm a suitable hosting location and infrastructure before onboarding, subject to availability. We document the agreed operator, data flow, and location for your review. Location is one part of that review, not a guarantee of compliance.
Do you train on our prompts or documents?
No. Your prompts, documents, and outputs are never used to train or fine-tune any model, ours or anyone else's, and they are not forwarded to a public model API. This is both contractual and architectural: there is no third-party model provider in the request path.
How long do you retain prompts?
The default posture is not to retain prompt or completion content at all. We keep operational metadata needed for billing, rate limiting, and abuse prevention. Your Data Processing Addendum states the exact retention period and any logging you explicitly opt into.
Are you SOC 2 certified?
Not today, and we would rather tell you now than in week three of diligence. rewired.io is an owner-operated infrastructure business. What we provide is a documented control set, a named operator, a defined data flow, and a hardware location your auditor can review, plus a signed DPA. If your procurement gate requires a completed certification, tell us early and we will say so directly.
What uptime do you commit to?
Planned shared access, if offered, runs on a commercially reasonable effort basis with no contractual uptime figure. Support scope and any response commitments are agreed before onboarding, separately from uptime. A measured uptime SLA with defined remedies is available on dedicated hardware engagements, written into an order form. Details are in the Terms of Service.
Who has administrative access to the hardware?
A named, deliberately small set of operators, over authenticated channels only, with physical access controlled by us. We will tell you who those people are during a security review. That is one of the advantages of buying from an operator rather than a support queue.
What if we want to leave, or you go out of business?
Because the API is OpenAI-compatible, leaving is a one-env-var change back to whatever you were using before. That is deliberate, so you are never locked in. You can export any data we hold for you before deletion, and custom LoRA weights trained on your data are yours to take with you.
Which models are proposed, and under what licence?
Proposed models, subject to workload evaluation and hosting availability. Model, hardware, and access are confirmed before onboarding. Mistral Small 4 requires a larger deployment assessment.
- Qwen3.8-27B (Apache 2.0)
- Ministral 3 14B Instruct (Apache 2.0)
- Mistral Small 4 (Apache 2.0)
We confirm the applicable model licence and deployment configuration before you build on it.
Can we send protected health information?
We don’t currently support workloads involving protected health information. Please don’t send patient data. Healthcare requirements can be discussed for future engagements.
Which subprocessors are involved?
We keep the list short on purpose: a hosting and CDN provider for this website, a form-delivery provider for contact submissions, and an email provider. Inference itself introduces no additional subprocessor. We provide the current named list on request and give customers advance notice of material changes.
Start here
Tell us what legal blocked. We'll show the private path.
Share your industry, who needs chat vs. who needs API access, and your compliance constraints. We respond with tier fit, availability, and onboarding steps, usually within one business day.
Direct line
Loading the contact form. You can also email hello@rewired.io.
We reply within one business day.