New features, improvements, and fixes — updated as we ship them.
New
New guide: Token Economics — the unit economics of AI
Every LLM call is metered in tokens, which means every AI feature has a cost of goods sold. Our new guide at /token-economics breaks down the five metrics that define your token economics and the four levers that actually lower the bill.
What token economics is, in one quotable definition — and why falling token prices still mean rising token bills.
The five core metrics: cost per request, per workflow, per feature, per customer, and failure waste.
The optimisation playbook: right-size the model, trim the context, eliminate failure waste, attribute then alert.
How every metric in the guide maps to something PromptLayer measures for you out of the box.
New
Never miss a spike again — alerts now find you, starting with Slack
Your AI observability just got a pager. Stop refreshing dashboards: the moment spend or failures move, PromptLayer tells you — in Slack, by email, or straight into your own tooling via webhook.
Add to Slack in one click: connect your workspace, pick a channel, and alerts land where your team already lives — no config files, no copy-pasted tokens.
Your thresholds, your rules: tell us the line — spend, failure rate, or failure spend — and we will ping you the instant it is crossed.
Dial in the volume per project: a minimum-severity floor means you choose between everything, warnings-and-up, or critical-only. Loud when it matters, silent when it does not.
Webhooks open the door to everywhere else: every alert ships as a clean JSON payload, ready for Zapier or any tooling you already run.
Zero alert fatigue by design: an ongoing condition notifies once and re-arms after a quiet window — you get the signal, never the spam.
All of it powered by the detection engine you already trust: failure-rate spikes, latency outliers, high-cost requests and workflows, repeated timeouts, and missing pricing.
Improved
Runtime Intelligence, redesigned
See what your AI is costing you, why it moved, and what to do about it — all at a glance.
New summary strip across the top: AI spend with its trend and a responsive sparkline, forecast, potential savings, and an efficiency score — readable at a glance.
"What changed" card surfaces the top driver of your spend with a one-click path to the underlying evidence.
Quantified recommendation cards: each shows the action, a confidence track, estimated monthly savings, and a model-swap suggestion where one applies.
Ownership & allocation card attributes spend to its top feature, customer, or team.
Data quality & confidence panel with at-a-glance rings, so you know how much to trust the numbers.
Refreshed marketing homepage to match.
New
Runtime Intelligence dashboard
A new project view that explains what changed in your AI spend, why, and what to do next.
Executive KPIs for spend (with trend sparkline), forecast, potential savings, and an efficiency score.
Spend causality and rules-based recommendations — no prompt content is analyzed.
Cost allocation by feature, customer, and team.
Runtime health at a glance: success and failure rate, latency, retries, and tool errors.
Workflow explorer plus a collapsible model-level supporting-evidence table.
