Home / Platform / Deployment

In your cloud. With your models.

Celara runs in your own cloud account and calls the models you already pay for, with your keys. Your data stays where your security team already reviewed it, and nothing goes to another AI vendor.

How you run it

Your account today. Ours when you want it.

Start in your own account. Hosted is on the way for teams that would rather we run it. Our engineers can join you either way.

Bring your own account

Available now

Celara installs into your cloud account and uses your model keys.

  • Your data stays in your account
  • Your models, on your provider bill
  • No token or invocation charges

Hosted

Coming soon

We run Celara for you on our platform, in a US region.

  • Each organization isolated at the database level
  • Runs in a US region
  • The same constitution, memory and evidence

Forward deployed engineers

If you need them

Our engineers work alongside your team when you want the help.

  • Tailor a solution to your systems
  • Build one the catalog doesn’t have yet
  • Get your first solutions live on your rules
Bring your own account

What runs where. All of it in your account.

Celara’s services, your data and your model keys all live in your cloud account. Agents call the model providers you chose. Nothing goes to another AI vendor.

YOUR CLOUD ACCOUNT your-company · your-region CELARA SERVICES Constitution & assurance Fabric, the governed runtime Memory Origin & Studio Your installed solutions MCP server and API for your developers YOUR DATA CRM · ERP · document stores Memory and knowledge graph Evidence and exceptions ledgers YOUR MODEL KEYS Model gateway with failover Allowed models set by tier Cost and tokens on every run YOUR PROVIDERS The models you chose called with your keys billed to you directly OTHER AI VENDORS Nothing sent no data, no prompts YOUR CLOUD ACCOUNT CELARA SERVICES Constitution & assurance Fabric, the governed runtime Memory Origin & Studio Your installed solutions YOUR DATA CRM, ERP, documents Memory Ledgers YOUR MODEL KEYS Model gateway allowed models by tier YOUR PROVIDERS Models you chose called with your keys OTHER AI VENDORS Nothing sent no data, no prompts
Bring your own account. Account and region names are placeholders.

Your data stays put

Memory, documents and ledgers live in your account. Celara connects to your systems of record where they are.

Your models, your bill

Agents call the providers you chose with your keys. You set which models each tier may use, and the gateway fails over between them.

One security review

One platform replaces a stack of AI tools, so your team reviews one deployment instead of one per vendor.

Security & data handling

Every decision on record. Every person’s data under control.

The same constitution that governs your solutions keeps the record. Personal data can be redacted before it is stored, and erased on request with proof.

Evidence ledger · append-onlyrecording
09:41:07
Contract reviewLiability cap within playbook · LLM judge
pass
09:41:12
Vendor onboardingApproval above threshold · checklist
awaiting approval
09:42:30
Policy answersCites a handbook clause · regex
pass
09:43:02
KYC reviewException granted · approver ≠ requester
excused
09:43:55
MemoryErasure request · 0 records remain
verified
Illustrative entriesExample figures
  • Append-only evidence. Every check against your constitution is written to a ledger that can’t be edited after the fact.
  • Exceptions with two names. An exception is granted by someone other than the person who asked for it, and both are recorded.
  • PII redaction. Switch it on and personal data is redacted before it reaches memory.
  • Erasure with proof. An erasure request removes a person’s data from memory and verifies that zero records remain.
  • People in the loop. The steps you choose wait for a named approver before anything happens.

Constitution & assurance

Articles, bindings, the evidence ledger and the exceptions ledger.

Explore the constitution →

Memory

Facts with their sources, PII redaction and verified erasure.

Explore memory →

Fabric

Model allowlists by tier, approval steps, and cost and errors for every run.

Explore Fabric →
Pricing model

You pay for what you deploy. Never for tokens.

When Celara runs in your account, the price follows what you put to work and who uses it. Talk to us for a quote on your scope.

By what you deploy

Agents, tools, skills and graphs

The pieces you install from the marketplace or build in Studio.

By user

The people who use it

Everyone who works with your solutions, asks them for help or approves their steps.

By connection

The systems you connect

Each system of record your solutions read from or write to, such as a CRM, an ERP or a document store.

You pay your model provider directly.
no token chargesno invocation chargesno published price list
FAQ

Questions security teams ask.

Does any of our data leave our account?

Celara runs inside your cloud account. Agents send prompts only to the model providers you chose, with your keys, and connect only to the systems you connect. Nothing goes to another AI vendor.

Which models can we use?

The models you have keys for. You decide which models each tier of agent may use, and the model gateway fails over between them when a provider is down.

When will Hosted be available?

Hosted is coming soon. We will run Celara for you on our platform in a US region, with each organization’s data isolated at the database level. Tell us if you want to hear when it opens.

Do we need forward deployed engineers?

Only if you want them. Most teams install from the marketplace or describe what they need in Studio. Our engineers can tailor a solution to your systems, or build one the catalog doesn’t have yet, alongside your team.

How is it priced?

By what you deploy (agents, tools, skills and graphs), by user and by connection. You pay your model provider directly, so we never charge for tokens or invocations. Talk to us for a quote.

How do we erase someone’s data?

Send an erasure request. Celara removes the person’s data from memory and verifies that zero records remain. You can also switch on PII redaction so personal data is redacted before it is stored.

Can we see what every agent did?

Yes. Every run is recorded with its cost, tokens, latency and errors, and every check against your constitution is written to the append-only evidence ledger.

Get started

Run it in your own account.

Bring one process. We’ll install the solution in your account, on your rules, in a working session and show you the evidence it produces.