Today we are excited to share the general availability of Tool Registry, a new functionality within Revenium that brings full-stack economic attribution to agentic deployments. Tool Registry maps every external call and review step back to the agent decision path that initiated it, enabling teams to demonstrate concrete ROI and economic impact to financial stakeholders.
Why “Token Tracking” Breaks the Moment Agents Hit the Real World
Most AI cost reporting stops at model usage, but tokens are often the smallest line item in enterprise workflows.
In production, the real spend is usually driven by what agents do beyond the model. This includes paid APIs, third-party services, internal systems, and human review steps.
That is why “token dashboards” create a false sense of control. They show the cheapest part of the workflow while the expensive actions happen elsewhere, commonly outside engineering’s line of sight.
Those costs can add up to tens of dollars per outcome while token spend stays in the cents, and they often arrive later as opaque vendor invoices with no clean link back to the agent decision path that triggered them.
That's exactly the gap Tool Registry was built to close.
This is what that looks like in practice. Token costs are the visible ‘waterline,’ while tool costs make up the bulk of spend.
The cost iceberg: token spend is visible, but tool costs make up the bulk of agent spend.
What the Tool Registry is
This release expands Revenium's economic observability beyond tokens to include registered tools, auto-discovered tools, and human-in-the-loop work, all metered and attributed consistently.
Tool Registry turns “AI cost” into a full-stack economic ledger, where every dollar can be traced to who (agent), what (tool), why (decision path), and where (customer, product, org).
This includes:
- External REST APIs
- MCP servers
- SaaS platforms
- Internal compute functions
- AI services beyond LLM completions
- Human review time
How Tool Registry Works
If an agent chooses to pull a credit report, the cost is not just “Equifax spend this month.”
The cost is tied to the exact trace that invoked it, to the agent and workflow that chose it, to the customer (or subscriber credential) and product context, and is visible next to token costs in the same system of record.
That is what makes Tool Registry useful for both engineering and finance. It creates economic accountability at the trace level, not after-the-fact invoice reconciliation.
Tool Registry provides teams with:
Universal cost registration that registers any tool with per-call pricing or tiered pricing models. This models real vendor contracts, including volume discounts.
Full-stack attribution which links every tool call to the agent decision that caused it, allowing teams to answer the question of“Which workflow drove last week’s over-budget alert?”.
Auto-discovery for unknown tools that detects and surfaces previously unregistered tools found at runtime. This is useful for large agent systems where tools proliferate quickly.
Circuit breakers that halt execution when per-trace or per-workflow cost ceilings are met, preventing runaway tool spend before it becomes a month-end surprise.
Trace-level cost distribution: spot outliers and enforce thresholds before costs spike.
Trace-level cost distribution: spot outliers and enforce thresholds before costs spike.
Unified analytics to see token costs and tool costs in one dashboard, and break down spend by org, product, agent, workflow, and customer.
Tool costs over time: understand spikes, drivers, and where spend is coming from by each tool.
Human-in-the-loop is a first-class cost event
Tool Registry treats human review as a tracked-cost event within the same trace.
That matters because many teams measure “automation” without measuring the economic result. If a high-volume workflow starts at 35% human review and drops to 12% over six months, that is a hard-dollar ROI story.
When human time is metered next to AI and tool costs, you can show:
- How the mix changes over time
- Where review is still required
- Whether automation is actually reducing cost per outcome
Why Tool Registry exists (and what it replaces)
Tool Registry fills a gap that adjacent tools were not built to address.
AI observability tools track latency, errors, and token counts, but can’t tell you whether a workflow was economically worthwhile.
Cloud FinOps tools roll up spend by team and project, but cannot trace costs across vendor boundaries back to a specific agent decision.
Tool Registry operates at the trace level, connecting every external call to the agent path that initiated it, across any vendor.
Get Started with Tool Registry
Tool Registry is generally available today in all versions for Revenium. Sign up for free to get started or learn more about Tool Registry on our docs site.
Top comments (0)