Show HN: Agent framework that generates its own topology and evolves at runtime

Source: Dev.to

Forget those “AI agent” demos that fall apart after five steps. This new framework, Hive, isn’t playing around. The team claims to build autonomous, self‑improving agents that actually work in production—without you hard‑coding a single workflow. They spent four years in the trenches of ERP automation for construction, so they’ve seen firsthand how brittle and useless most “AI” tools are for real business.

Core insight: Accountants want the ledger reconciled while they sleep, not to chat with a bot. They want services, not tools.

The “How It Works”

You don’t design workflows, draw flowcharts, or even define agent interactions. Instead, you chat with a coding agent inside Hive and tell it your high‑level business goal—e.g., “reconcile all invoices in this folder and log discrepancies.”

What Hive does next is wild:

• It generates its own topology – the coding agent builds the entire agent system, creating a node graph (think dynamic workflow) and the code to connect everything.
• It self‑heals & evolves – when (not if) the agent hits a snag in the messy real world, Hive captures the failure data, the coding agent evolves the graph, fixes the broken parts, and redeploys. It’s a self‑improving loop.
• LLM‑agnostic – GPT‑4o, Claude Opus, Google Gemini, Mistral, even your local Ollama models. Hive supports >100 LLMs via LiteLLM, so you’re not locked into one vendor.
• Human‑in‑the‑loop – built‑in nodes for human oversight, credential management, and real‑time monitoring.

This isn’t a “tool”; it’s a system that builds and manages its own tools to achieve your defined outcome.

The “Lazy Strategy”

“Lazy” here means smart. You let AI do the grunt work of building and debugging.

1. Define outcomes, not steps – stop thinking “do X, then Y.” Think “I need Z to happen.” Let Hive figure out the X and Y.

2. Use WSL or Git Bash on Windows – avoid Command Prompt or PowerShell; the Linux subsystem gives you a smoother experience.

3. Clone the repo – it’s open‑source under Apache 2.0.

   git clone https://github.com/adenhq/hive.git

4. Plug in your preferred LLM – set your API keys, or run a local Ollama model and point Hive at it.

5. Start with the examples – explore exports/ and examples/templates/. Don’t try to build a complex agent from scratch; learn the “Hive way” of defining outcomes.

6. Join their Discord – the community is active and looking for contributors.

The “lazy” part is letting the AI scaffold, debug, and evolve complex workflows while you focus on goals and oversight.

The Reality Check

Hold your horses—this isn’t a magic button, and the team is clear about that.

• Agents are still hard – “self‑improving” is a massive claim; it doesn’t mean “set it and forget it.” Early‑stage agents need monitoring, guidance, and refinement.
• Not for simple tasks – Hive is overkill for one‑off scripts or trivial API calls. It targets production‑grade, multi‑agent workflows.
• Developer‑focused – you need comfort with code, Python, and the usual software‑development pain. The coding agent helps, but you remain the orchestrator.
• Garbage In, Garbage Out – vague or contradictory high‑level goals produce equally vague agents. Precise outcome definition is a skill you’ll need to hone.
• Failure is expected – automatic failure recovery is built in, but you’ll still decide how gracefully and how often you intervene.

The Verdict

YES, absolutely—if you’re serious about building robust, autonomous business processes.

If you’re a developer or a team burned by brittle agent frameworks and need production‑grade automation that handles messy real‑world data, Hive is one of the most promising approaches I’ve seen. The promise of self‑evolving, self‑healing agents that construct their own topology from a high‑level goal is exactly what we need to escape the current agent‑framework hell.

• If you just want a quick script or a simple chatbot, this isn’t it.
• If you’re ready to tackle the hard problems of autonomous AI that actually works, clone the repo, dig into the examples, and try to break it. That’s how you learn—and possibly build something truly game‑changing.

🛠️ The “AI Automation” Experiment

I’m documenting my journey of building a fully automated content system.

• Project Start: Feb 2026
• Current Day: Day 4
• Goal: To build a sustainable passive‑income stream through AI‑driven content creation.

(More updates to follow…)