Skip to main content

Agent Integration

Automate the operation you actually need

MatterSpace for constrained generation, DLMS for dynamic system learning and modeling, CDE for causal discovery from data, ACI for post-launch model adaptation. REST, SDK, MCP, and CLI are access methods. The value is in what each product delivers.

Base: https://api.vareon.com · Auth: X-API-Key · Format: JSON · Spec: OpenAPI 3.1

MatterSpace

The Universal Generation Engine for Science and Engineering

Use MatterSpace when the job is generating viable candidates from target properties and constraints — materials, molecules, circuits, algorithms, or biological candidates ready for ranking and testing.

MethodEndpointDescription
POST/v1/generateSubmit constraints, get candidates
POST/v1/validateCheck candidate against domain rules
GET/v1/generations/{id}Retrieve generation results
Auth

API key via X-API-Key header

DLMS

Dynamic Learning and Modeling System

Use DLMS when the job is learning how systems behave, evolve, and respond to change — dynamic system learning, trajectories, and behavior modeling through the Model Discovery Engine (MDE).

MethodEndpointDescription
POST/v1/learnSubmit observations, learn system dynamics
POST/v1/modelBuild system model from learned dynamics
GET/v1/trajectories/{id}Retrieve trajectory predictions
POST/v1/discoverRun MDE discovery on dynamic data
Auth

API key via X-API-Key header

CDE

The Causal Dynamics Engine for Science and Engineering

Use CDE when the job is turning observational data into equations, causal structure, or system models — interpretable results your team can review, validate, and deploy.

MethodEndpointDescription
POST/v1/discoverSubmit episodes, get causal theory
POST/v1/sessionsCreate persistent discovery session
POST/v1/campaignsMulti-run research program
GET/v1/runs/{run_id}/resultRetrieve run artifacts
GET/v1/claims/{claim_id}Inspect ledger-backed claim
Auth

API key via X-API-Key header, scoped to organization and project

Auto-discovery

/.well-known/ai-plugin.json and /.well-known/mcp.json

ACI

Continual Learning After Deployment

Use ACI when the job is keeping deployed AI adapting after launch: per-tenant updates in shared services, device-local memory and erase, or bounded adaptation on edge hardware.

MethodEndpointDescription
POST/v1/bindAttach scoped learned state
POST/v1/unbindRemove learned contribution exactly
POST/v1/adaptIncorporate new data in bounded time
POST/v1/constrainApply typed safety constraint
POST/v1/rollbackRestore prior state deterministically
GET/v1/audit/{scope_id}Retrieve signed audit trail
Auth

API key via X-API-Key header, scoped to organization and tenant

Auto-discovery

/.well-known/mcp.json

Choose the protocol after you choose the job

ProtocolUse whenFormatAuth
REST APIDefault. Any language, any framework.JSON over HTTPS. OpenAPI 3.1 specification available at each product's /docs/api.X-API-Key header
Python SDKNotebooks, batch jobs, CI pipelines.Typed client: pip install cde-sdk / pip install aci-sdk. Sync and async clients.API key passed to client constructor or VAREON_API_KEY env var
MCPAgent hosts that implement Model Context Protocol.Standard MCP. Tools discovered via /.well-known/mcp.json. Parameters, return shapes, and side effects declared per tool.Session-scoped credentials. Same policy enforcement as REST.
CLITerminal automation, CI/CD, operator workflows.cde <command> / aci <command>. Machine-readable JSON output with --format json.Config profiles per environment. API key in profile or env var.

Operational controls where the boundary requires them

Approval gates

Require explicit approval for sensitive operations instead of leaving execution policy to prompt text.

Budget ceilings

Bound compute, API calls, and parallel runs so autonomous workflows stay within resource limits.

Safety boundaries

Declare forbidden actions, state limits, or route restrictions where the host workflow requires them.

Audit trail

Keep signed lineage for teams that need to inspect exactly what an automated workflow did.

Start integrating

Start from your workflow, then pick the product surface and protocol. Every endpoint is documented with request and response schemas.