RatioCore Book a demo
RatioCore
Book a demo

Deployment and control

Run agents on your terms.

Run RatioCore on your own servers, in your cloud or on AMD Cloud. Decide which systems it connects to, which models it uses and who approves changes, so it fits the rules your organization already follows.

YOUR ENVIRONMENT
Agent runtime
PostgreSQL data
Trace records
Backups
Optional local inference
Outbound: only the channels, connectors and models you approve

Operate from chat

Check on your deployment from a chat.

In this Telegram example, an operator asks for status and finds out who approved an action, without opening a console. Every channel you enable is an outbound path, reviewed like any other connector.

Own the environment

Three decisions you keep.

Deployment is more than hosting. Decide where RatioCore runs, who can authorize sensitive actions and how operators find out what changed.

Self-host where you choose

Evaluate private cloud or on-premises options against your data, model, network and recovery requirements. Details depend on the workload and operating environment.

  • On premises or private cloud
  • Customer-managed infrastructure
  • Planned around each workload

Name an owner per action

Use roles, permissions and action-level approvals. RatioCore rechecks authority when the action is about to run, so approval stays tied to the intended operation.

  • Role-based permissions
  • Action-level approvals
  • Authority rechecked at dispatch

Know what changed

Inspect run history, approvals and recorded effects. If an external change has an unknown outcome, reconcile it before anyone retries.

  • Run history and approvals
  • Recorded external effects
  • Reconcile before any retry

Sovereign deployment options

Make the boundary an explicit choice.

Some components can stay inside your hosting boundary. Others can carry data out of it. List both before you enable anything.

  • External modelsConnect an approved external model only where policy permits. Review what context is sent with each request.
  • Messaging channelsChannels can move data outside the hosting boundary. Review their access, retention and network paths first.
  • APIs and connectorsEach connector reaches a system with its own data path. Decide which integrations to enable, and with what scope.

Self-hosting is a deployment option. Sovereignty depends on the full configuration. External models, channels and connectors add their own data paths.

Deployment reference for technical evaluators

What the current reference uses.

The current reference configuration is a starting point for evaluation. Reference configurations are not universal deployment guarantees.

01

PostgreSQL 17 data store

The current reference stores application data in PostgreSQL 17. Confirm where the database, trace records and backups will reside.

02

Vector and lexical search

Retrieval uses pgvector alongside lexical search, so agents can find relevant documents and records within the deployment.

03

Separate model routes

Different workflows can use different model routes. Model routes never carry authority to execute actions, wherever the model runs.

04

Identity and recovery

Review identity integration, isolation, backup and recovery against your specific requirements during evaluation.

What to verify in an enterprise pilot

Prove it on one real workflow.

  • Run a permitted action and a denied action
  • Confirm who could authorize each, and what the record shows
  • Trace one sensitive record from intake through to audit
  • Change a review rule and check ownership and audit hold
  • Map storage, inference, backups and every outbound connection
  • Interrupt a run and agree how operators reconcile it

FAQ

Deployment questions.

Still deciding? Bring your questions to a walkthrough.

Can RatioCore run fully on premises?

RatioCore can be self-hosted on customer-managed infrastructure, on premises or in a private cloud. Whether a workflow stays fully inside your boundary depends on the full configuration, including where models run and which channels and connectors you enable.

What does the current reference deployment use?

The current reference uses PostgreSQL 17 with pgvector and lexical search. Review model placement, connector scope, backup and recovery, and identity requirements during evaluation. Reference configurations are not universal deployment guarantees.

Can we use an external model?

Yes, where your policy permits. You can evaluate local inference for restricted workflows and connect an approved external model for others. Review what context is sent with each request. Model routes never carry authority to execute actions.

What happens if a run is interrupted mid-action?

If an external change has an uncertain outcome, it must be reconciled before anyone retries it. Operators can inspect run history, approvals and recorded effects to see what happened. Agree the reconciliation process during your pilot.

See it on your business

Which function would you reinvent first?

Bring a process your team runs today. In a demo we show how agents could run it, where your people stay in charge, and what a focused pilot could prove.

See it on your workflowBook a demo