Enterprise AI with clear controls.

VertexGraph integrates with existing enterprise data and agent platforms through Core APIs, Dynamic APIs, and approved HTTP or MCP tools. Use authorized data to answer questions and execute tasks, and bring VertexGraph experts and Skills into existing workflows.

System architecture

Logical overview
  • Synchronous calls
  • Asynchronous job delivery
  • Data and storage access

Scroll horizontally to explore the full map.

Business entry points
Access and governance
Knowledge and AI runtime
Async jobs and processing
Data and output storage
Employee workspace
Enterprise applications
Enterprise agent platform
Virtual employees
sk-admin-webAdmin console
Enterprise data platform
Enterprise sources
JWT · API KeyIdentity verification
Dynamic APIExpert and Skill access
Core APIUser and knowledge management
sk-backIdentity, permissions and orchestration
Knowledge access tokens
Third-party registration
HTTP · MCPApproved tools
Multilingual query planning
sk-searchGraph and vector retrieval
Reranking and evidence
Virtual experts
sk-skillSkill runtime
Configured model services
Usage and run monitoring
sk-task APITask registration
Outbox Relay
Redis Streams
Task Worker
Scheduler
Parsing and extraction
Graph and vector indexing
Source files and attachments
MySQLBusiness state and Outbox
MySQLUser sessions and attachments
Neo4jGraph index
MilvusVector index
Published deliverables
Task and run records

Access and traceability

Enforce user and knowledge scopes at retrieval and execution. Manage published tools and retain API calls, runs, and output status for investigation.

Deployment options

Run on company-managed servers or in a private cloud. An isolated deployment requires internal model services, prepared model files, compatible hardware, and document rendering resources.

Service and data flows

  • Synchronous calls
  • Asynchronous job delivery
  • Data and storage access

Scroll horizontally to explore the full map.

Enterprise applications
sk-admin-webAdmin console
sk-backIdentity, permissions, experts, and APIs
AIConfigured model services
sk-searchAuthorized RAG and reranking
sk-skill API
sk-task API
MySQLBusiness state and Outbox
Scheduler
Skill Worker
Task Worker
Outbox Relay
Neo4jGraph index
Controlled file and deliverable storage
MilvusVector index
Redis Streams
HTTP / MCPApproved tools

Knowledge ingestion

The backend commits the knowledge change and an outbox event together. The relay publishes it to Redis Streams. Task Workers parse and index the document and update the recorded state.

Evidence retrieval

Search checks the user’s knowledge scope, plans across document languages, combines graph and vector results, and reranks evidence before returning it to the expert or Skill.

Skill execution and delivery

The Skill API hands work to a separate runtime. The Worker uses the authorized knowledge and approved tools, records run state, and publishes eligible outputs to shared storage.

Integrate through two API surfaces

Dynamic APIs

Agent platforms can call published VertexGraph experts and Skills, track task status, read responses, and retrieve published outputs. Each endpoint has its own access key.

POST /dynamic/api/{api_path}
Content-Type: multipart/form-data

userid=10001
command=Returns policy
rag=true
stream=true

Core APIs

Data platforms can synchronize users, manage authorized directories and documents, and search for matching passages grouped by document. Call Core APIs from your server with a Core API key.

POST /api/v1/knowledge/search
Content-Type: application/json

{
  "user_id": 10001,
  "question": "Returns policy",
  "node_ids": [
    12
  ]
}

For example, an agent requests an operations report. VertexGraph retrieves metrics through an approved data tool, drafts and publishes the report using a template, and hands it back to the existing approval and distribution workflow.

Put your first AI expert to work.

Log in