Hi HN, we’re building AxonFlow for teams running LLMs or agents in real production systems.
Once agent workflows move past demos, failures are rarely model issues. They tend to show up as execution problems during real runs.
Short 2-minute technical demo showing execution control and auditability in practice:
https://youtu.be/FNgnESo9RtI
AxonFlow is a self-hosted, source-available (BSL 1.1) control plane that sits inline in the execution path and governs LLM calls, tool calls, retries, approvals, and policy enforcement step by step. It does not replace your orchestrator and can run alongside LangChain, CrewAI, or custom systems.
The problems we focus on are usually discovered only after going to production:
- retries that accidentally repeat side effects
- partial failures mid-workflow
- permissions that differ per step
- limited ability to inspect or intervene during execution
This is not aimed at early demos or hobby projects. It’s for teams already operating under real production constraints.
GitHub:
https://github.com/getaxonflow/axonflow
Docs:
https://docs.getaxonflow.com
I’d value feedback from folks running LLM or agent workflows in production.
Once agent workflows move past demos, failures are rarely model issues. They tend to show up as execution problems during real runs.
Short 2-minute technical demo showing execution control and auditability in practice: https://youtu.be/FNgnESo9RtI
AxonFlow is a self-hosted, source-available (BSL 1.1) control plane that sits inline in the execution path and governs LLM calls, tool calls, retries, approvals, and policy enforcement step by step. It does not replace your orchestrator and can run alongside LangChain, CrewAI, or custom systems.
The problems we focus on are usually discovered only after going to production: - retries that accidentally repeat side effects - partial failures mid-workflow - permissions that differ per step - limited ability to inspect or intervene during execution
This is not aimed at early demos or hobby projects. It’s for teams already operating under real production constraints.
GitHub: https://github.com/getaxonflow/axonflow
Docs: https://docs.getaxonflow.com
I’d value feedback from folks running LLM or agent workflows in production.