AgentNexus Docs

AgentNexus is a full-stack AI agent management platform that lets you create, train, deploy, and manage intelligent conversational AI agents.

Core features include: a visual agent builder, knowledge base management, multi-channel deployment, conversation history tracking, and the ancli CLI tool.

Whether you're a solo developer or an enterprise team, AgentNexus helps you launch a professional AI agent in minutes.

1. Sign Up: Visit AgentNexus and click 'Get Started Free'. Register with email, Google, or GitHub.

2. Create Agent: In the Dashboard, click 'Create Agent' and set the name, system prompt, and LLM model.

3. Add Knowledge: Go to agent details and upload documents, paste URLs, or type text into the knowledge base.

4. Test: Use the Sandbox chat to test conversations with your agent and fine-tune responses.

5. Deploy: Use the Deploy feature to launch your agent to Fly.io or Railway with one click.

Each agent includes: name, description, system prompt, LLM model selection, temperature, and knowledge base config.

Supported LLM models: Connect to 100+ models via OpenRouter, including GPT-4o, Claude 3.5, Gemini, Kimi K2.5, and more.

Agent configuration can be changed at any time — changes take effect immediately without redeployment.

Dashboard: Create, edit, delete, and clone agents. CLI: ancli agents list / create / get / update / delete.

The Knowledge Base lets your agent answer questions based on your content, instead of relying on generic AI knowledge.

Three knowledge sources are supported: text input, URL scraping, and file upload (PDF, TXT, CSV, etc.).

Under the hood, pgvector is used for semantic retrieval (RAG), ensuring accurate and contextual answers.

CLI: ancli knowledge list / add --text / add --url / add --file / delete.

All agent conversations are automatically saved, including user messages and AI responses.

In the Dashboard, you can filter by agent, view full conversation history, or delete specific conversations.

Conversation data is used to analyze agent performance and optimize prompts.

CLI: ancli conversations list / list --agent <id> / get / delete.

AgentNexus supports one-click deployment to Fly.io and Railway cloud platforms.

Deployment flow: Select target platform → Configure region and model → Auto-generate Dockerfile and config → Deploy.

After deployment, health check endpoints are automatically generated, with custom domain and SSL support.

CLI: ancli deploy generate (preview config) / ancli deploy fly --token <FLY_API_TOKEN>.

ancli is the AgentNexus CLI tool (current version: v0.1.0) that supports all platform features from the command line.

Install: cd packages/cli && npm install && npx tsc

Auth: Copy your token from Dashboard Settings → Developer / CLI, then run: ancli auth login --token <token>.

Core command groups: auth (authentication), agents, knowledge, conversations, deploy, config.

All commands support -o json output format for AI Agent and script usage. Exit codes: 0=success, 1=error, 2=bad args, 3=auth failed, 4=not found.

Full manual: doc/manual/ancli_user_manual.md

Account: View and manage account information (email, username, User ID).

Developer / CLI: Get your CLI token and view quick start commands here.

Plans: Free (50 messages/mo) / Standard ($49/mo, 3,000 messages) / Pro ($199/mo, 10,000 messages).

Appearance: Supports light/dark/system theme switching.

Language: Switch between English, Chinese, and Spanish interface.

AgentNexus provides a complete REST API. All Dashboard features are available via API calls.

Authentication: Requests must include a valid Supabase Auth session cookie or API token.

Key endpoints:

All data transmission uses TLS 1.3 encryption.

User authentication is powered by Supabase Auth, supporting Google/GitHub OAuth and email/password login.

Session tokens are stored in httpOnly cookies to prevent XSS attacks.

Agent and conversation data is isolated by organization for multi-tenant security.

CLI tokens are served via a server-side API on the Settings page only, never exposed in client-side JavaScript.