Launching today
Agentium is a TypeScript-native AI Agent runtime for building production agent systems without graph sprawl or Python glue. Create agents, coordinate teams, manage memory, call tools, automate browsers, run workflows, and trace every model call, tool hop, and memory write from one modular Node.js stack. Start with core, add only the modules you need, and ship observable AI products that feel like real application code, not fragile orchestration scripts.
Hey Product Hunt! I’m excited to launch Agentium today.
Agentium is a TypeScript-native AI agent runtime for developers building real agent systems inside Node.js products.
We built it to solve a problem we kept running into: once agents move beyond demos, the stack gets messy fast. You end up with graph wiring, Python glue, custom memory adapters, tool calls, browser automation, queues, and observability all living in separate places.
With Agentium, you get one modular runtime for:
Agents and multi-agent teams
Create typed agents and coordinate them without fragile orchestration scripts.
Workflows and tools
Define stateful workflows, call APIs, connect tools, and automate browser actions.
Memory
Keep session history, user facts, summaries, decisions, learnings, and retrieval context in one runtime.
Observability
Trace every model call, tool hop, browser action, and memory write.
Here’s how simple it is to create an agent:
Start with @agentium/core, then add only what you need: browser automation, queues, evals, observability, admin APIs, transports, or Studio.
We’d love feedback from builders here:
What integrations should we prioritize?
What would make Agentium easier to adopt in your stack?
Where do agent frameworks still feel too painful today?