Revenium 2.17.0 is about deepening the economic intelligence customers use every day. Most of this release strengthens capabilities teams already run in production, making the numbers more trustworthy, the controls more flexible, and the risk detection more automatic.
Here are the biggest things in this release.
1) Unified Guardrails bring notifications, hard limits, and inspection into one place
Spend notifications and hard limits now live together as Guardrails. From one place, teams can manage the two side by side, review rule history, and give teammates read-only inspection access so reviewers can see the rules without being able to change them.
Guardrails also hand off cleanly from AI by Employee. You can start a Guardrail prefilled from a filtered employee view, and show developers your own message when an enforced limit blocks a request. When a rule fires, developers see guidance from the owner rather than a generic error.
This update provides:
- One place to manage spend notifications and hard limits
- Rule history and read-only inspection for reviewers
- Guardrails prefilled from a filtered AI-by-Employee view
- Custom developer-facing messages when a rule blocks a request
2) Smarter cost alerts that catch problems before the bill does
AI Insights now watches for two spend-risk patterns on its own.
The first is when cost per call is climbing faster than usage, a signal that a model change, prompt regression, or new tool has quietly made every call more expensive.
The second is when a single API key, subscriber, product, or agent breaks from its own historical spending pattern, which surfaces anomalies at the entity level rather than only in the aggregate.
Both patterns show up before the provider bill does, so teams find out about drift while there is still time to act on it.
3) Automatic spike explanations you can actually defend
When a day's AI spend jumps, Revenium now explains the cause on its own from the AI Assistant view. It identifies which people and which usage drove the spike against their own normal pattern, and sets that against what the team produced that day.
A high-spend day used to require a manual investigation across dashboards. Now leaders see the same day with context already attached. They can tell a genuine efficiency problem from a productive burst and act on the real cause instead of chasing a number.
4) Billed vs metered clarity, everywhere it matters
Every analytics surface now labels whether a figure is Billed, meaning an actual provider invoice, or Metered, meaning usage Revenium observes directly. Alongside that, a new provider spend reconciliation view compares billed and metered totals with clear coverage-health states and cleaner handling of data gaps or unusually high ratios.
Together, these remove one of the most common sources of confusion in AI cost work. Finance and engineering get a shared vocabulary for the same number, and they can see at a glance whether instrumentation is keeping up with actual provider spend.
5) Richer AI spend analysis by employee
AI by Employee goes considerably deeper in this release. Teams can filter employee-level usage by model tier, provider, and vendor, compare individuals against team benchmarks and cost per million tokens, choose the columns that matter, and export results to CSV.
That gives leaders the shape they actually need for reviews. Costly usage patterns become easier to identify, fair comparisons become easier to make, and the analysis can travel into whatever workflow the team already uses to talk about performance.
Also in this release
GitLab integration. GitLab-based teams can connect a group from the Integrations page and optionally narrow the connection to a specific project. A direct, manageable route into Revenium without broadening access beyond the scope a team wants.
Read-only role. A view-only role across the workspace lets finance partners, executives, and auditors see dashboards, budgets, and settings without any risk of accidental changes. More of the organization can be brought into the picture safely.
GitHub user-mapping improvements. Existing user mappings can now be edited, and multiple GitHub accounts can be mapped to a single corporate email. Developers who use more than one GitHub identity keep per-employee cost attribution accurate instead of fragmenting across accounts.
Useful previews for empty accounts. Analytics and coding-assistant dashboards can show representative sample data before a customer's own data arrives, so new users can understand the product's value before setup is complete.
Alongside these headline features, 2.17.0 also ships a redesigned Tasks page, a new Provider Comparison chart template, coding-assistant activity included by default with a per-team opt-out, improved AI Insights attribution, metering logs search, more reliable metering delivery in the Python, Node.js, and Go integrations, and a range of accuracy, performance, and API compatibility improvements. For the full technical details on everything that shipped, the changelog has the complete rundown.
Try it out
Revenium 2.17.0 is live in production.
If you're already using Revenium, the fastest places to start are unified Guardrails, the new AI Insights cost-risk alerts, and the billed-versus-metered view. Together, they make it easier to trust the numbers, catch problems earlier, and put real controls in front of AI spend.
For the full list of improvements, fixes, and technical details, read the public changelog at docs.revenium.io/get-started/changelog.
