Axiom

Spotlight + AI is here ✨ - Investigate anomalies in seconds—no guesswork, no endless query iteration. Select a region on a chart → Spotlight compares it against baseline behavior → Get a clear, trustworthy summary of what’s different Learn more ↓

I got this working, here's how 1. Created agent-first clis for viewing logs in (@getsentry and @AxiomFM) 2. Sentry sends a webhook to backend on new error 3. Backend triggers a github action 4. Github action triggers codex (authenticated via chatgpt sub) 5. Github action creates PR when codex is done

@itzdgofficial @ericzakariasson 🫡

@ericzakariasson @AxiomFM can we have one

actually super excited about cursor plugins. so easy to package your best practises and distribute it. combining datadog (logs) + databricks (analytics) + slack (human conversations) is so nice, especially in cloud agents where it can do a lot of async troubleshooting

🧪*

Learn more → axi.sh/LoL7jCk

From the Changelog: Write evaluations skill for AI agents! :test_tube: └ Install in Claude Code, Cursor, Amp, Codex, or any compatible agent └ Generates eval suites from natural language prompts └ Supports exact match, LLM-as-judge, online evals, and more Turn your AI assistant into an eval suite author.

#axiom" target="_blank" rel="nofollow noopener">supabase.com/docs/guides/te…

⚡️ Supabase Log Drains are now available on Pro and they support Axiom! Your Supabase logs - Postgres, Auth, Storage, Edge Functions, Realtime - now flow directly into Axiom. Set up in minutes ↓

🚨 Announcing our second Pre-Flight class of The TinyFish Accelerator 1500+ requests, 35 public demos later, we have selected our next 10 builders! These builders unlocked free API access, engineering support, and weekly mentorship. See what they shipped👇🧵

From the Changelog: Online evaluations for AI engineering are here! 🧠 └ Score your AI capability's outputs against live production traffic in real-time └ Reference-free scorers work without ground truth — boolean, numeric, or LLM-as-judge └ Full OTel tracing links every score back to the generation span that produced it Stop flying blind after deployment. Learn more → axi.sh/hoxfAZm

Offline evals → Online evals → User feedback. The full loop 🔄 └ Know if you’re ready to ship └ Know if what you shipped is working └ Know what users think └ Each piece closes a different blind spot Start scoring production traffic → axi.sh/QmaTEeY

Every score links back to the trace 🔍 └ Click through to the exact input, model call, and output └ No guessing what went wrong—you’re looking at it └ Production failures become offline test cases └ Your test coverage grows from real traffic, not hypotheticals

Offline evals catch regressions before you ship. User feedback catches problems after users notice. But what about the gap in between? └ Long-tail inputs you never tested for └ Format drift after a model provider update └ Subtle quality degradation nobody reports That gap is now closed 👇

@chronark Hi, no, a bug we’re trying to work out where the account cleanup l doesn’t happen right with third parties and so we revert to contact sales or support. Still trying to track down the issue exactly. Shared internally we’ll manually delete and cleanup. Apologies for the issue!

really? @AxiomFM we just moved to our own grafana to consolidate and I cannot even close my account before talking to sales?

@thisismahmoud @thallescomumh @Railway let's make it happen? :)

@thallescomumh @AxiomFM @Railway What's your use case? 👀

please please please @AxiomFM add @Railway as a integration please thats all I ask for please please please 🙏🙏🙏

It's a bad week to be a slow piece of code in the @AxiomFM codebase

Learn more → axi.sh/MIUSXvp

From the Changelog: Query metrics skill for AI agents! 🤖 └ Turn your AI coding agents into metrics exploration experts └ Works with Claude Code, Cursor, Amp, Codex, and more └ Discover metrics, explore tags, and execute queries using natural language Just ask your agent "Query the CPU usage metric" and let it handle the rest.

The new landing page for evlog.dev is live After a month of building, here's where we're at: Frameworks & runtimes: • First-class Nuxt, Next.js & TanStack Start • Hono, Cloudflare Workers, Bun, Node & more Drain to anywhere: • @axiomfm, @posthog, @sentry, @BetterStackHQ • OTLP, custom adapters • Batching, retries, fan-out Core features: • Wide events with full context • Structured errors with why/fix suggestions • Smart sampling (head + tail) • AI-ready output • Client to server transport