What we include
Tools that a FinOps or platform team could use as the central control point for AI traffic and AI spend: AI gateways (in the request path), AI FinOps platforms (reporting, allocation, sometimes enforcement) and LLM observability tools with cost tracking. Native cloud controls (Amazon Bedrock, Azure API Management, Google Apigee/Vertex AI) are included because many teams use them instead of a separate gateway.
How we collect data
- We read vendor documentation, pricing pages, GitHub repositories and official announcements. Every non-trivial cell links to its source.
- If we cannot confirm a fact in a public source, we write unknown. We never guess. "unknown" does not mean the feature is missing.
- "no" means the vendor's documentation shows no such feature (for example, a log export that is not in FOCUS format).
- "n/a" means the criterion does not apply, for example request-blocking budgets for a reporting tool that is not in the request path.
- Every tool has a last checked date. Current data: 2026-10-20. We plan to re-check every tool at least monthly and record changes in the weekly newsletter.
How we score
The FinOps score measures how well a tool supports cost control and allocation. Six criteria are scored:
- Hard budgets (team / key / user)
- Rate & spend limits
- Cost attribution / chargeback tags
- FOCUS export
- EU data residency option
- Public pricing
Public pricing is "yes" when the plan that includes the scored features has a public price, "partial" when only some prices are public (for example, budgets need an Enterprise plan or an add-on with no public price), and "no" when pricing is quote-only.
Each criterion gives 2 points for "yes", 1 for "partial" and 0 for "no" or "unknown". Criteria marked "n/a" are left out. The score is the share of points on the applicable criteria. Open source, hosting, provider count and caching are shown but not scored, because the right answer depends on your situation.
Limits of the score: it counts documented capabilities, not quality, performance or ease of use. Compare tools within the same category. A FinOps platform and a gateway do different jobs.
Current data quality
25 tools, 11 tracked fields each. Of 275 cells, 202 are sourced facts, 48 are unknown and 25 are not applicable.
Neutrality and conflicts of interest
- No paid ranking. Ever. Sponsors and verified vendors cannot change a score, a cell value or the sort order.
- Sponsorships will be clearly labelled "Sponsored" and kept apart from the comparison data.
- A paid "verified profile" means the vendor reviewed its facts and supplied evidence. Unverified facts stay as they are; the score uses the same rules.
- No affiliate links at launch. If we add any later, we will label them and list them here.
- Corrections: anyone (including vendors) can send a correction with a public source. We publish the change and the date.
Data use
The full dataset is downloadable as JSON and CSV. Please cite "The Token Ledger" and the check date.