Start an AI product without assembling the foundation by hand.
Choose a production-ready backend, then add agent runtime, RAG, tools, memory, guardrails, and model access as real modules. Workspai keeps the project connected to the wider workspace from the first change.
Step 1 · Choose the foundation
A real backend before an AI demo
Start with FastAPI or NestJS using the same governed project flow as the rest of your software. No separate AI-only island is required.
Step 2 · Add what the product needs
Composable AI and agent capabilities
Each module has a version, dependency contract, compatibility data, and quality evidence. Install one capability or compose a complete AI backend.
Agent Runtime
Agent runtime scaffolding for tools, memory, execution state, retries, and observable AI workflows
AI Assistant
Provider-agnostic...
AI Guardrails
AI input/output safety policy, PII redaction hooks, and model boundary checks
LLM Gateway
Provider-neutral LLM gateway with routing, fallback, budgets, and usage telemetry for AI products
Prompt Ops
Prompt versioning, rollout, evaluation, rollback, and audit operations for AI products
RAG Pipeline
RAG pipeline scaffolding for ingestion, embeddings, retrieval, citations, and evaluation-ready AI search
Tool Registry
Tool registry, schemas, permissions, and audit-safe invocation policy for AI products
Vector Store
Vector store abstraction for pgvector/Qdrant-style indexes, similarity search, and embedding metadata
Need the graph, Doctor, skills, and verified repair too?
Workspai maps projects and dependencies, prepares agent-ready context and skills, finds blockers, and verifies fixes across the workspace. Your new project stays connected to that shared workspace from the start.
Explore Workspace Intelligence