Self hosted
Run the agent inside infrastructure your organization owns and operates.
Best for: Maximum control, internal governance and custom operating requirements.
Deploy Agents
Run the agent where your controls belong.
Use the same Silvana agent logic across infrastructure models that place data, secrets, signing and operations in different boundaries.
Same agent logic. Different control boundaries.
Four ways to run
Choose the boundary that fits the mandate.
The same agent can move through different runtime models as security, scale, confidentiality and operating requirements change.
Cloud deployment
Deploy the agent in your cloud account or approved cloud project.
Best for: Scalable infrastructure with familiar cloud controls and team owned operations.
Silvana managed
Run the agent in a Silvana managed environment with a defined shared responsibility model.
Best for: Faster launch, lower operational overhead and supported production workflows.
Trusted Execution Environment
Run supported agent logic and sensitive material inside an isolated runtime.
Best for: Workflows that require stronger confidentiality around strategy, data or signing.
How it connects
Keep strategy logic separate from settlement infrastructure.
A Silvana agent does not need to rebuild the full Canton integration for every workflow. It uses SDK or gRPC clients to observe state, prepare actions, verify transactions, sign inside the approved boundary and route execution through Silvana services.
Agent logic
Applies the strategy, mandate, policy and workflow rules.
SDK or gRPC clients
Connects the runtime to Silvana services for orderbook, settlement, ledger and supported data flows.
Verification and signing
Reviews prepared transactions and signs inside the approved environment.
Silvana execution services
Coordinate supported order, transaction and settlement workflows.
Canton settlement
Provides the privacy and final settlement foundation for asset state changes.
Decision framework
Choose by control, not convenience alone.
A deployment model is a control decision. Before choosing the runtime, decide where the important boundaries should sit.
Signing boundary
Where are prepared transactions reviewed and signed?
Data boundary
Where can private strategy inputs, market context and operational state be processed?
Secret boundary
Where are credentials, keys, tokens and configuration stored?
Network boundary
Which services, models, data sources and internal systems can the agent reach?
Operating boundary
Who owns releases, monitoring, incident response, updates and recovery?
Governance boundary
Which logs, approvals, access rules and configuration records must be retained?
One continuous path
Build locally. Choose the runtime. Operate with intent.
Build
Use the SDK guide to install tooling, onboard the agent and complete the first working flow.
Validate
Use the Playground and supported nonproduction environments to inspect behavior and signing flow.
Deploy
Choose the runtime model, document the responsibility boundary and launch the approved version.
Deploy with intent
Put the agent in the right boundary.
Start with a working agent, choose where logic and authorization belong, then operate the runtime with clear ownership from the first deployment.