Built to fit existing institutional infrastructure.

Chat-native interface

Compass agents operate inside the client's existing chat platform. Analysts, models, and execution flows share one channel, with threaded deliberation on disagreements, structured handoffs, and full attribution back to the audit log. Multiple agents can argue out a signal in front of the operator, with each verdict tagged by model identity, confidence, and reasoning.

The research surface is the workflow surface. There is no separate terminal to learn.

Model fleet

Compass orchestrates a heterogeneous fleet across frontier APIs and local inference, including Anthropic Claude, Google Gemini, OpenAI Codex, and Ollama-hosted open-weight models. Routing is per task. Model identity is recorded on every verdict. Local-only deployments are supported for clients with provider restrictions.

OMS and fund accounting

Bluewater integrates with order management systems and fund accounting platforms over standard transports. Position, P&L, and execution data flow in and out of the same audit trail as the rest of the platform. Reconciliation against book of record is continuous, not periodic. Connectors are built per client against documented OMS and accounting interfaces.

Markets and data

Native ingestion for prediction markets (Polymarket today), digital asset venues, and equity and news feeds. Position and price data persist for the lifetime of the deployment. On-chain execution is wallet-aware and routed through the same scheduler as scripted and AI-driven workflows.

Deployment and stack

Meridian is built on Microsoft SQL Server and Windows Server. Internal services communicate over TCP. Web-facing components run on IIS. The platform is designed for on-premises installation or private cloud deployment (Azure, AWS, or customer-managed virtualization) within institutional environments. Data residency, network posture, and audit retention are deployment-time decisions.

Security and access

Meridian deployments default to Windows Authentication against Active Directory for both application and database access, with role-based controls through SQL Server permissions and Compass-side ACLs. Enterprise SSO and secrets-management platforms are supported for client-managed deployments.

Data is encrypted at rest using SQL Server Transparent Data Encryption and Windows BitLocker; service-to-service traffic is encrypted in transit. For deployments with provider restrictions, Compass can run AI workloads on local Ollama-hosted models exclusively, with no third-party API calls or data egress. Audit retention is bounded only by SQL Server storage; standard institutional windows operate on the same infrastructure as live operations.

On the horizon

Active development continues across asset-class coverage, additional chat-platform support, and new detector tooling. Specifics available under NDA.

Inquiries

Detailed integration reference, supported systems, and deployment patterns are available under NDA. Contact sales@lambspring.ai.