reddit · SideProject · 09-26 19:50 46 分 ← 返回列表

I built a way to see your app, AI calls and errors in one trace

Solo founder here. When an AI feature breaks in prod, the error is in Sentry, the spans are somewhere else, and the model call is in a third tool. axonpush puts them together: point your existing OpenTelemetry exporter or Sentry DSN at it, no new SDK app spans, exceptions and model/tool calls on one trace; errors grouped into issues hand a failing run to Claude Code or Codex with the trace attached optional gateway for AI traffic: block or redact risky tool calls and PII, and cap sp

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：AI 应用生产环境故障排查时，错误、链路 span 和模型/工具调用分散在 Sentry、追踪系统和第三方工具中，难以关联到同一条 trace。

目标用户：AI 应用开发团队和独立开发者，尤其是已使用 OpenTelemetry 或 Sentry 的工程团队。

证据：When an AI feature breaks in prod, the error is in Sentry, the spans are somewhere else, and the model call is in a third tool. It's early, and I'm looking for the first few teams to try it. I'll set it up with you personally.

变现：按 trace 量或团队席位订阅收费；高级功能如 PII 脱敏、工具调用拦截、每个 agent 的花费上限作为增值付费点。

MVP：2 周内做一个轻量 MVP：接收现有 OpenTelemetry exporter 或 Sentry DSN，把 app span、异常和模型/工具调用合并展示在同一条 trace 上，支持基础错误分组和分享 trace 链接，暂不做复杂网关和策略引擎。

竞品：Sentry、Langfuse、LangSmith、Helicone、Braintrust、Arize Phoenix、Traceloop/OpenLLMetry

摘要：Solo founder 发布 axonpush，可将现有 OpenTelemetry exporter 或 Sentry DSN 接入，把 app spans、异常和模型/工具调用放到同一条 trace，并支持把失败运行交给 Claude Code/Codex。帖子在寻找首批试用团队，未展示收入数据或明确付费意愿，且 AI observability 赛道已有多个成熟竞品。