Building Emeralbase: An Early AI Agent Platform for Sales, Support & Lead Generation
AI chatbots are easy to demo but hard to turn into something genuinely useful for a business. Most tools stop at “upload docs → chatbot” and leave teams guessing about trust, usage, and outcomes.
I’ve been building Emeralbase, an early-stage platform for creating custom AI agents that can act as a sales agent, support agent, lead generator, or general knowledge assistant, depending on how you configure it.
👉 Early demo: Emeralbase
⚠️ Runs on free-tier infrastructure — expect rate limits and occasional slow downs.
What Emeralbase Can Act As
Instead of a single “chatbot,” Emeralbase is designed around roles:
Lead generation agent
Capture leads naturally during conversations using custom fields (email, company size, use case, etc.).
Sales agent
If you have a product, the agent can answer pricing and feature questions, qualify interest, and guide users toward conversion.
Support agent
Trained on your documentation, FAQs, and internal content to handle common support questions.
General knowledge agent
Useful for internal teams or knowledge bases where users need fast answers from shared documents.
The same core system adapts based on how you train and configure it.
What Emeralbase Does Today
These are the features that currently exist (no roadmap promises here):
🧪 Playground
Test and refine agent behavior before deploying it live.
📚 Multi-Source Training
Train agents using:
- Websites
- PDFs
- Text
- Documents
🧠 Multiple Model Support
Experiment with different AI models without rebuilding the agent.
💬 Chat Logs
Full visibility into conversations for debugging and insight.
📊 Chat Analytics (with Geography)
Track usage patterns and see where your users are coming from geographically.
🎯 Lead Generation via Natural Conversations
Leads are captured inside the chat flow, not through disruptive forms.
🧾 Captured Leads Dashboard
View and manage collected leads directly from the dashboard.
🚀 Deployment Options
Deploy agents using:
- A widget
- An iframe
This makes it easy to embed on almost any website.
What’s Still Coming (Upcoming)
Some features are planned but not built yet:
- Advanced document versioning and refresh workflows
- Better chunk inspection and debugging tools
- Deeper analytics and cost visibility
- Integrations (Slack, Intercom, Zendesk, etc.)
- More opinionated flows for sales vs support
The current focus is validating core use cases before expanding.
Why I’m Building This
This started as a side project because I wanted:
- More clarity around what AI agents actually know
- Better insight into real user conversations
- A clearer path from chat → lead → action
Instead of optimizing for “AI wow factor,” the goal is to build something that fits into real business workflows.
Who This Is Best For (Right Now)
- SaaS teams experimenting with AI sales or support
- Small businesses testing AI-driven lead capture
- Developers building internal knowledge agents
- Teams comparing different AI models on real data
Try the Early Version
⚠️ Free-tier deployment — rough edges and rate limits are expected.
If you try it, I’d love feedback on:
- Which role (sales, support, lead gen, knowledge) feels most valuable
- What feels confusing or missing
- Whether the analytics and logs are useful
Honest feedback — even critical — is extremely helpful at this stage.
Thanks for reading 🙏

Top comments (0)