01CONTROL
Plans you can govern
Preview and approve a plan or an individual dispatch, require roles and ordering, and gate every default-deny tool call.
SCOPE
Enterprise AI Agent Delivery · Rev. 2026.07COMPANY
YuanASI · ShenzhenCAPABILITY
Consulting / Development / IntegrationEnterprise AI delivery · advisory, development & integration
Put AI agents into
real business workflows.[01]
From use-case assessment to production handoff, YuanASI delivers measurable, maintainable AI agent systems that fit your stack.
→ Book a 30-min business assessmentFree AI diagnosis · live agent report30-minute initial assessment · no technical brief required
DELIVERY PATH · FROM BUSINESS PROBLEM TO PRODUCTION HANDOFFDWG. DELIVERY-01
- 01Assess
- 02Validate POC
- 03Integrate
- 04Deploy
- 05Handover
10+ more
02 / PRODUCTION EVIDENCEDWG. OMA-SERIES · SHEET 02 / 04
OPEN SOURCE RUNTIME
open-multi-agentControl the plan.
Prove every run.
Open Multi-Agent (OMA) generates and runs task DAGs inside TypeScript / Node.js applications, with plan and per-task approvals, execution receipts, traces and a Run Viewer, EvalSets, and task-level checkpoints. Your organization chooses and controls the models, tools, credentials, and deployment.
RELEASE SNAPSHOT · v1.13.0
↗A real run replay, not a concept diagram
02EVIDENCE
Runs you can inspect
Stable run IDs, execution receipts, and traces feed a local Run Viewer or export to OpenTelemetry.
03EVALUATION
Quality you can measure
Versioned EvalSets, scorers, offline reports, CI gates, and online sampling support regression decisions before and after release.
03 / DELIVERY SYSTEMSHEET 03 / 04
Enterprise AI agent systems
built to keep running—
with measurable business value.
Enter at the stage your organization is in today. Every stage has defined inputs, engineering outputs, and acceptance evidence.
STAGE01
DWG D-01
SCALE 1:1
Scenario validation & acceptance baseline
Prove the case before expanding the investment.
- Business outcomes, data boundaries, and non-AI constraints
- POC plus a versioned EvalSet and reference cases
- Acceptance report and Go / No-Go recommendation
STAGE02
DWG D-02
SCALE 1:1
System integration & production control
Connect agents to existing systems and encode execution boundaries.
- Dynamic or explicit workflows plus enterprise APIs and data
- Plan approval, tool policy, timeouts, budgets, and task checkpoints
- Trace / Run Viewer, regression evaluation, and CI gates
STAGE03
DWG D-03
SCALE 1:1
Private deployment & team handoff
Run in your environment, then transfer operational ownership to your team.
- Node.js / TypeScript engineering assets and private deployment
- OpenTelemetry integration, runbooks, and troubleshooting paths
- Source, tests, SOPs, and team training
INTAKE / NEXT STEP
Enter at the stage your organization is in today. Every stage has defined inputs, engineering outputs, and acceptance evidence.
04 / INTAKE · SHEET 04 / 04
Ready to ship AI agents?
Start with a 30-minute call.
→ Book a free 30-minute intakeCALFree AI readiness diagnosis · live agent reportAGTExplore open-multi-agent on GitHub↗
Studio
Nanshan District, Shenzhen
Software Industry Base, Building 1 · A701-20