Integrations
Works alongside your stack. Replaces none of it.
Vera connects to your existing system of record over REST, makes verified decisions, and acts through open interfaces. Open standards end to end — Docker, REST, MCP — so there is no vendor lock-in at any layer, including the model.
Data in
Connect the systems you already run.
Honest labels: what's in production today versus what's on the integration roadmap.
Field-service management
AvailableJobs, technicians, certifications, SLAs and schedules pulled over REST from your FSM system. Vera maintains its own normalized view — no schema migration on your side.
TMS / ERP
AvailableShipments, routes, status and timing from your transport management or ERP system via REST. Vera works alongside the current TMS — no rip-and-replace.
Telematics & vehicles
AvailableVehicle positions, availability and driver-hour state, so working-time and routing rules are checked against reality, not yesterday's plan.
EDI
RoadmapStructured carrier and partner messages (status, delays, bookings) feeding the same decision pipeline.
Email ingestion
RoadmapFree-text disruption notices from carriers and customers, translated by the language layer into structured facts — the same mechanism used for chat input today.
Decisions out
Consume Vera as a product — or as a component.
REST API
POST /v1/reason for verified decisions, /v1/plans/evaluate for SAT/UNSAT with unsat cores, /v1/rules/parse for rule translation, plus memory and audit reads. Scoped API keys per integration.
MCP server
vera_reason, vera_evaluate_plan, vera_parse_rule, vera_get_rules, vera_query_memory — exposed as Model Context Protocol tools so your agent frameworks can delegate constraint checking to Vera.
Notifications & actions
Customer ETA notifications, dispatch updates and write-backs to your operational system, executed by the agent at the autonomy level you configure.
Models
Model-agnostic by design.
The language layer sits behind a provider-agnostic abstraction with schema validation. The solver — and therefore decision integrity — is unaffected by model choice.
Self-hosted EU models
Open-weight European models (Mistral family, Teuken class) via Ollama or vLLM. The sovereign configuration: nothing leaves your deployment.
Hosted API models
Any OpenAI-compatible endpoint your organization approves. Only prompt text reaches the endpoint — never your operational stores.
Bring your own
Connect your own approved model endpoint. Cutover requires passing our extraction evaluation suite at parity or better — model swaps are a config change, not a re-architecture.
Integration is days, not months.
No schema migration, no consultants. Tell us what you run and we'll map the connection in the first call.