Forge — AI Product Development Workbench
"Everything is a conversation" · dual-core Ontology + Session (internal R&D)
Internal R&D Platform (Not a Commercial SaaS)
Forge is currently positioned as a Lurus internal R&D tool (ontology-driven requirements management + API Gateway demo) — it is not a commercial product offered for sale. Access is invite-only beta, and the API is still evolving. For questions or partnership inquiries, contact business@lurus.cn.
What Is Forge?
Lurus Forge is a development workbench for AI product teams, built on the core philosophy that "everything is a conversation" — product requirements are discussed through Session conversations, features are implemented by AI Agents (PM/Architect/Code), and knowledge is visualized through a product Ontology.
Under the hood, the Kova engine provides WAL persistence for Agent tasks, so work resumes seamlessly even if execution is interrupted.
A tree structure that organizes a product's user stories, architecture, tech stack, and design specs — static, structured knowledge.
Each product discussion is packed into a Session — a dynamic timeline carrying conversations, decisions, and Agent output.
From Requirements to PR, All in One Visual Structure
Shipped and planned capabilities are shown side by side, with status labels marked honestly.
Product Ontology
Tree-based management of user stories / architecture / tech stack / design specs, with every dimension visualized side by side; decisions made in conversation automatically update the Ontology.
Conversation-Driven Development
Ask “What is the user story for this feature?” → the PM Agent analyzes and generates it. Every decision is linked to its conversation context, so you can trace back why it was decided that way.
WAL Decision Replay
Built on the Kova engine WAL, every conversation step and decision is persisted — traceable, locatable, and ready for replay-based review.
Planned / In-Development Capabilities
Three-tier dependency management that goes beyond Renovate/Dependabot: Patch auto-merges (zero manual work); Minor offers one-click approval cards; Major triggers a conversational review (the AI analyzes the semantic business impact of breaking changes).
Drag-and-drop three-stage builder: Trigger (Webhook / schedule / API request) → Process (LLM call / RAG retrieval / tool call) → Action (API callback / email notification / database write).
RAG knowledge base: document import (PDF/Word/Markdown/web pages), automatic chunking (preserving semantic integrity), vector indexing (automatic embeddings for semantic search), and update sync (automatic re-indexing when documents change).
Call statistics (volume / latency / tokens), quality scoring (user feedback + automated evaluation), cost analysis (by feature / time), and alerts (automatic notification on abnormal call volume or quality drops).
Prompt Engineering Workbench
| Feature | Description |
|---|---|
| Prompt Editor | Syntax highlighting, variable insertion, version management |
| A/B Testing | Compare output quality of different prompts for the same input |
| Model Comparison | Compare how the same prompt performs across different models |
| Batch Testing | Import a test set and evaluate in bulk |
| Version History | Every edit is saved as a version automatically, with rollback anytime |
What Teams Do on Forge
- AI SupportVisually Build a Support AgentManage the knowledge base, monitor service quality
- Content ModerationDrag-and-Drop Moderation FlowsSet rules, keep optimizing
- RecommendationsConfigure a Recommendation AgentA/B test different strategies
- Document QAImport Docs to Build a Knowledge BaseDeploy a Q&A Agent
Tech Stack
| Layer | Technology |
|---|---|
| Frontend | TypeScript + React (Turbo monorepo) |
| AI Engine | Lurus API (multi-model support) |
| Agent Runtime | Kova (persistent execution) |
| Vector Store | Qdrant / Chroma |
| Deployment | Kubernetes (ArgoCD) |
Invite-Only Beta
Forge is currently in invite-only beta. It's a good fit for teams that:
Are integrating, or plan to integrate, AI features into their product
Need visual prompt management and testing tools
Want to lower the development and operations cost of AI features
Apply for Beta Access
Email business@lurus.cn, noting your team size and the pain points you hope to solve.
More Resources
- Lurus API — Learn about the underlying AI capabilities
- Kova — The persistent Agent execution engine
- MemX — AI intelligent memory management
- Unified Authentication — Forge login / team permissions / SSO federation are all built on this