Build production-ready AI agents with knowledge, tools, and real-time collaboration.
Website
·
Issues
⚠️ Early Development - Niceyup is currently in early development. Features may change, and some functionality may be incomplete. We're actively working on improvements and welcome feedback!
Niceyup is an AI agent platform built for teams that want reliable, production-ready assistants, not just chatbots. It combines agent behavior controls, model and embedding configuration, tool integrations, and knowledge retrieval in one collaborative workspace.
With Niceyup, you can connect documents, Q&A data, databases, and external MCP tools, then run real-time conversations with grounded responses and source attribution. From internal copilots to support and operations workflows, it gives you the infrastructure to build, govern, and scale useful AI teammates.
Build production-ready AI agents with:
- Behavior and instruction controls (system prompts and response style)
- Multi-model support (OpenAI, Anthropic, Google, OpenAI-compatible, and gateway providers)
- Fully customizable model parameters (temperature, top-p, reasoning effort, and more)
- Tool use and function calling (native and external capabilities)
- Embedding model configuration (for retrieval quality and grounding)
- Real-time chat experience (streaming responses and responsive message updates)
- Conversation controls (regenerate, resend, and stop message generation)
- File attachments in chat (context-aware conversations with uploaded files)
- Persistent history (conversation context across sessions)
- Shared workspace usage (collaborative, team-based conversations)
- Behavior configuration (instructions, guardrails, and response direction)
- Model and embedding tuning (quality, latency, and cost balancing)
- Knowledge retrieval controls (top-k, indexing, and search strategy)
- Tool and integration enablement (MCP, providers, and runtime capabilities)
Connect and ingest various data sources:
- Text Documents - Upload PDFs, Word docs, markdown, and more
- Question & Answer Pairs - Structured Q&A entries for targeted answers
- Source Explorer - Organize sources and folders at the team level
- Automatic Ingestion - Background processing powered by Trigger.dev
- Agent-level knowledge base configuration (per-agent retrieval behavior)
- Automatic chunking and embedding pipeline (optimized for semantic retrieval)
- Source attribution in responses (grounded answers with references)
- Retrieval-Augmented Generation (RAG) for higher answer quality
- Reindex workflows (refresh indexed data as sources evolve)
- Connect any MCP server (bring your own tools and capabilities)
- Integrate model providers (including OpenAI-compatible endpoints)
- Connect vector store providers (currently Upstash)
- Centralized integration management (providers, MCP servers, and connections)
- Organizations - Multi-tenant workspaces with isolated data
- Teams - Group agents, sources, and workflows by team
- Role-Based Access Control - Fine-grained permissions and governance
- Invitations - Invite, assign, and manage members
- Shared Resources - Collaborate on agents and knowledge bases
We are working on the documentation to get started with Niceyup for local development: https://docs.niceyup.com/local-development
- Monorepo
- pnpm + Turborepo
- React
- TypeScript
- Next.js 16
- Fastify 5
- Drizzle ORM
- PostgreSQL
- Shadcn/ui
- TailwindCSS
- Vercel AI SDK
- LangChain
- Better Auth
- Vercel (Web)
- Railway (API)
- Neon (Database)
- Upstash (Redis)
- Cloudflare R2 (Storage)
- Trigger.dev (Background jobs)
- Resend (Transactional & Marketing)
- GitHub Actions (CI/CD)
- GitHub OAuth (Social authentication)
- Polar (Payment processing)
- Novu (Multi-channel notifications)
- OpenAI (AI models)
- Anthropic (AI models)
- Google (AI models)
This project is licensed under the AGPL-3.0 for non-commercial use.
For commercial use or deployments requiring a setup fee, please contact us for a commercial license at hello@niceyup.team.
By using this software, you agree to the terms of the license.