Sagad Console
Supervisor UI for queues, approvals, conversations, QA context, agent performance, knowledge inventory, integrations, and settings.
SagadOS by JD SYS
Open-source AI-native BPO infrastructure.
Sagad OS coordinates inboxes, CRMs, knowledge bases, approvals, QA, observability, prompt work, and tool execution through a supervised Agent Studio layer.
Open-source v0.1.0AI-native BPO platformSelf-hostableHuman approval firstAgent Studio backendChatwoot + Twenty adapters
Live open-source previewv0.1.0
What it is
A control layer for AI-assisted customer operations.
Sagad OS is for BPO, contact-center, support, and sales teams that need AI help without losing human control, context, or auditability.
Agent Studio
FastAPI + LangGraph orchestration layer for typed state, retrieval, draft generation, approval gates, traces, and server-side adapters.
Adapter Boundary
Chatwoot, Twenty CRM, Uptime Kuma, MCP, and client tools stay behind controlled backend adapters instead of browser-direct calls.
Operating loop
From message to approved action.
Intake, context, draft, supervisor review, execution, and audit stay inside one inspectable workflow.
- Step 01
Customer message enters Chatwoot or another channel adapter.
- Step 02
Agent Studio creates typed conversation state.
- Step 03
Knowledge, SOP, QA, CRM, and tool context are retrieved.
- Step 04
AI drafts a response or tool plan with trace metadata.
- Step 05
AI Ops Supervisor reviews, approves, edits, rejects, or takes over.
- Step 06
Approved reply or external action runs through a server-side adapter.
- Step 07
Audit, trace, and handoff state stay inspectable.
Desktop Preview
Current Sagad Console screens.
Supervisor Console
Command Center
Click image to inspect
Desktop preview for queue health, SLA risk, approval load, channel readiness, and knowledge readiness.
Approval Worklist
Attention Queue
Click image to inspect
Desktop preview for conversations that need approval, escalation, failed send handling, or human takeover.
HITL Review
Conversation Review
Click image to inspect
Desktop preview for AI draft review, decision trail, KB/SOP/QA context, and approval control.
Adapter Catalog
Tool Readiness
Click image to inspect
Desktop preview for MCP/CRM adapter readiness, owner visibility, and approval-route status.
Planned Features
Prompt engineering is a first-class product surface.
The next layer is not just more connectors. It is better prompt control, evaluation, retrieval, QA, and supervisor visibility around every AI-assisted action.
Prompt Engineering workspace
Prompt versioning and evaluation notes
Knowledge base and SOP retrieval
QA scoring and compliance review
Agent performance dashboards
Connector registry
Persistent audit events
MCP tools behind approval gates
Release Notes
Out now as an open-source preview.
The repo is public and usable for technical review. Production hardening remains explicit: auth, tenant isolation, persistent audit storage, backups, and deployment runbooks still need to mature before high-risk customer data goes through it.
Note 01
Repository is public at rdkrstn/sagad-os.
Note 02
Current version is 0.1.0.
Note 03
The console runs from v1/ and Agent Studio runs from agent-studio/.
Note 04
The project is out as an open-source preview, but not production-hardened yet.
Stack Signals
Built around supervised agent operations.
Next.jsTypeScriptTailwind CSSFastAPIPythonLangGraphLangChainLangSmithChatwootTwenty CRMUptime KumaDockerGitHub ActionsHITL approvals