Enterprise
Bring your own model: keys, endpoints, contracts
Point Ciao at the model relationships you already trust — your own provider keys, OpenAI-compatible endpoints and zero-retention contracts — instead of inheriting someone else's vendor choice.
Own-LLM on Ciao means your organization chooses the model provider behind its workspaces: bring your own keys, connect OpenAI-compatible endpoints, and keep inference under zero-retention contracts. Unlike platforms hard-wired to one vendor, Ciao runs a multi-provider model ladder with fallback, so your AI SDLC does not depend on a single model company — and customer code is never used to train models.
Published 2026-07-03 · Last updated 2026-07-03
Your model contracts are part of your risk posture
By now, most enterprises have done the hard work once: negotiated data-handling terms with a model provider, reviewed retention and training clauses, and cleared a specific vendor through security and legal. The last thing that team wants is a development platform that quietly routes company code through a different model relationship with terms nobody reviewed.
This is a real gap in the AI tooling market. Many products treat the model as an internal implementation detail — you get whichever vendor the platform picked, under whatever terms the platform negotiated for itself. For an individual developer that is a convenience. For an enterprise, it is an unreviewed sub-processor sitting in the middle of the software supply chain.
Ciao treats model choice as an enterprise decision. You can bring your own provider keys, connect OpenAI-compatible endpoints, and align inference with the zero-retention contracts your legal team has already blessed — while the platform's own multi-provider model ladder with fallback keeps delivery resilient when any single vendor degrades.
There is a procurement-efficiency argument here too. Every model relationship a platform introduces is a vendor assessment your team has to run — retention terms, training clauses, incident posture. Routing AI development through provider agreements you have already assessed means the new platform inherits finished work instead of generating new reviews, which is often the difference between a quarter and a year in an enterprise rollout timeline.
What Ciao provides
- Your own provider keys — Route model usage through keys your organization controls, so consumption, terms and vendor relationships stay under agreements you negotiated — configured with the enterprise team during onboarding.
- OpenAI-compatible endpoints — Connect endpoints that speak the OpenAI-compatible API surface, which is how most enterprise model gateways, proxies and approved-vendor setups are already exposed internally.
- Zero-retention inference — Inference runs under zero-retention model contracts, and customer code is not used to train models — the two clauses your review will look for first.
- A multi-provider ladder with fallback — Ciao's model ladder spans multiple providers with fallback, reducing dependency on any single model vendor — an availability property and a negotiating position at the same time.
- Audit across the loop — The append-only audit trail records prompts, merges, deploys and admin actions, so model-assisted work is attributable end to end regardless of which provider served the tokens.
- Governance independent of the model — Guardrails applies plain-English policies and records human review on risky changes — the control layer does not weaken or change when the model behind it does.
How own-LLM setup works in practice
1. Map your current model estate
With the enterprise team, list the providers, gateways and contracts you already have — most organizations discover they already hold the right agreements for this.
2. Choose the routing posture
Decide which workspaces use your keys and endpoints, and where Ciao's own multi-provider ladder applies — many teams mix both, using their contracted vendor as primary.
3. Connect and verify
Keys and OpenAI-compatible endpoints are configured for the workspace, and your team verifies traffic lands where you expect before production work begins.
4. Put the terms in the contract
Zero-retention and no-training commitments are contractual language reviewed during procurement, not footnotes — your legal team reads the actual clauses.
5. Operate with fallback
If a provider degrades, the model ladder's fallback keeps delivery moving, and the audit trail keeps recording who did what across the change.
Verification and commercial notes
Own-LLM configuration is part of an enterprise engagement, scoped with the team rather than toggled self-serve, because the right setup depends on your gateways, contracts and network posture. The claims that matter here are contractual: customer code is not used to train models, and inference runs under zero-retention model contracts — both reviewable in the document set during procurement, alongside SOC 2 Type II reports under NDA. Serious production programs start at USD 10,000 per year. If your model strategy is still forming, the multi-provider ladder means you are not forced to decide today: you can start on Ciao's default posture and move workspaces to your own keys as your vendor strategy settles.
Two questions worth asking in any evaluation of this capability: what happens when a provider changes its terms, and what happens when one degrades. The multi-provider ladder exists so both answers stay boring — terms are contractual per provider, and fallback keeps delivery moving while the vendor conversation happens on your schedule rather than the outage's.
Model postures on Ciao
| Posture | Who holds the vendor relationship | Typical fit |
|---|---|---|
| Ciao model ladder | Ciao, across multiple providers with fallback | Default: resilient delivery without vendor lock |
| Your own keys | Your organization, under your negotiated terms | Enterprises with existing provider agreements |
| OpenAI-compatible endpoint | Your organization, via your gateway or proxy | Teams with an internal model gateway or approved-vendor list |
| Mixed | Both, per workspace | Primary on your contract, fallback for resilience |
Frequently asked questions
Which providers can we bring?
Any provider relationship you can expose through your own keys or an OpenAI-compatible endpoint, which covers the common enterprise setups including internal model gateways. The specifics of your estate are mapped with the enterprise team during scoping.
Is our code used to train models, whichever provider serves it?
No. Customer code is not used to train models, and inference runs under zero-retention model contracts. When you bring your own keys, your own negotiated terms govern that relationship as well.
What happens if our chosen provider has an outage?
Ciao runs a multi-provider model ladder with fallback, which reduces dependency on any single model vendor. How fallback interacts with your own-key posture is agreed during setup, so nothing routes anywhere your review has not approved.
Can we see which model activity produced which change?
The append-only audit trail records prompts, merges, deploys and admin actions, so AI-assisted work is attributable end to end. Guardrails additionally records human review on risky changes, which is usually the evidence auditors ask for.
Does governance depend on which model we use?
No. Guardrails, QA and Security operate on the code and the running application, not on trust in any particular model: policies apply in plain English, tests run before and after publish, and security findings are confirmed against the live app before they are flagged.