Introducing Respan AI Gateway.
The world’s first AI gateway with built-in observability, evals, and prompt optimization.
Access 1,000+ models through one endpoint, with the production layer teams need beyond routing.
your agent got worse in production and nobody told you
Respan now supports evals for full agent traces.
wire evaluators into your agents, track quality every day, and catch regressions before users start posting your failures on X.
Introducing Respan Evals 2.0: Automation.
Most AI production issues don’t show up as errors. They show up as answers that look valid but are wrong, incomplete, unsafe, or off-brand.
Respan now runs automated online evals on live production logs to catch silent quality failures
4/ Turn eval scores into alerts
Configure thresholds and trigger webhooks when quality drops.
Use them to notify your team, create review queues, or kick off incident workflows from actual eval results.
5/ Make evals part of production ops
Online evals should not live in a dashboard nobody checks.
Connect Respan to your monitoring, alerting, and incident response systems so AI quality becomes an operational signal, not a postmortem artifact.
Spent this week looking at AI eval tools. They're all the same tool.
LLM-as-a-judge, wrapped for model outputs. Then wrapped again for agent traces. That's the whole category. Nothing new under the hood.
The hard part is getting to your first eval that actually means something.
Just found this thing called logprobs, and OpenAI is hiding it on their newest models. It's the real % a model puts on each next word before it picks one.
So I asked 3 models how many R's are in "strawberry," had each one state its confidence, then checked the receipts.
All
Global monthly searches, AI coding tools:
Claude Code: 2.06M
GitHub Copilot: 721K
Cursor: 584K
OpenCode: 298K
Claude Code pulls more searches than Codex and Cursor combined. It's not close.
Thrilled to co-host this voice‑coding hackathon🚀
Join us Friday in SF for a night of building and experimenting together.
See you there!
@VoiceCursor@convex@photonhq
10 engineers × 4 CLI agents = 40 accounts, 40 invoices, 40 price tags.
We made it 1.
Claude Code, Codex, Gemini CLI, OpenCode, and more can now run through Respan Gateway with unified billing and discounted usage.
