cloud-finops-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Cloud FinOps knowledge by OptimNow. Two retrieval surfaces: REFERENCES (long-form provider/discipline files) and PLAYBOOKS (small named-pattern runbooks for specific waste patterns). REFERENCES: list_references() to discover, find_references(query=, domain=, capability=, phase=, persona=, maturity=, persona_primary_only=) to route a question - query= takes the user's own wording (English or French) and ranks by the questions each file declares it answers, the facets narrow by FinOps Capability/Phase - and get_reference(name=, section=) to fetch one body. Every listing entry carries approx_tokens: references run from about 2,000 to about 15,000 tokens. get_reference(name=, section="") returns one section instead of the whole file - section matches H2 and H3 headings case-insensitively and partially, and a miss returns the file's available_sections rather than the whole body. The Azure pattern catalogue (finops-azure-patterns) is an enumerated list, usually needed one section at a time; the AWS catalogue is one file per service family (finops-aws-patterns-compute, -storage, -databases, -networking, -other). PLAYBOOKS: list_playbooks() to discover, find_playbooks(query=, scope=, service=, waste_category=, confidence=) to route a symptom or a 'which of my X' question (query= takes the user's wording) or to narrow by pattern facets, get_playbook(name=) to fetch one body. A playbook serves 'how do I detect/fix this specific pattern' (zombie NAT, snapshot sprawl, idle ELB, etc.); a reference serves billing mechanics, commitment strategy, allocation methodology and cross-pattern reasoning. This connector has no access to the user's cloud accounts. For questions about the user's own resources ('which of my RIs are about to expire', 'which of our VMs run for nothing'), the matching playbook includes a detection query the user can run in their own account. For a specific waste or cost-fix question ('my NAT gateway processes 10TB to S3', 'should I delete these old snapshots'), find_playbooks returns the named runbook with its tested detection query. Runbooks exist for NAT gateways and VPC endpoints, expiring Savings Plans / RIs / reservations, snapshots, S3 lifecycle, idle or stopped VMs, orphaned disks and IPs, GPU and SageMaker sizing, and more. References carry billing mechanics, not current prices: any figure inside is illustrative and dated inline. Where a reference quotes a figure from an OptimToken guide with its date, get_live_figures(guide=) returns that guide's current figures with the snapshotDate to quote them by. For 'what changed recently' or 'what is coming' questions, get_reference(name='finops-whats-new') is a dated digest of the provider changes verified over the last 90 days, one section per month plus 'Scheduled changes': state its as-of date. Advisory questions - how much to commit, how to size, how to allocate or charge back - are served by a matching reference (finops-aws-commitments for Savings Plan sizing, finops-chargeback for recharge design), which carries the decision rules and cadence tables behind a defensible answer.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_referencesA | See what FinOps guidance is available: billing mechanics, commitment strategy, allocation and chargeback, AI cost management, and per-provider cost handbooks (AWS, Azure, GCP, OCI, Databricks, Snowflake, ...). |
| get_referenceA | Fetch the guidance on one FinOps topic - the billing mechanics, decision rules and worked examples behind a defensible answer - either whole or one section at a time. |
| find_referencesA | Find which guidance serves a FinOps question - how to commit, size, allocate, charge back, forecast, or govern cloud and AI spend. |
| list_playbooksA | See every ready-made runbook for finding and fixing cloud waste: idle, orphaned and overprovisioned resources, egress surprises, schedule blindness and AI/ML inefficiency across AWS, Azure and GCP. |
| get_playbookA | Fetch the step-by-step runbook for one specific waste pattern: symptoms, the detection queries to run, the fix, and the anti-pattern to avoid. |
| find_playbooksA | Finds runbooks for a suspected cloud-waste pattern, filterable by provider, service, waste category or detection confidence. Useful when the user asks how to detect or fix a specific waste pattern; each runbook includes a detection query the user can run in their own account. This connector has no access to the user's accounts. |
| get_live_figuresA | Get the current value of a figure a reference quotes with a date. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 7 tools
The set splits cleanly into two parallel families — reference/guidance tools (list_references, find_references, get_reference) and waste-pattern playbooks (list_playbooks, find_playbooks, get_playbook) — with list/find/get boundaries spelled out explicitly in each description. The only soft spot is get_live_figures, whose purpose (current values for figures quoted inside three named references) slightly overlaps with get_reference on those same guides.
Every tool follows the same verb_noun convention with a consistent singular/plural split: list_* for enumeration, find_* for ranked search, get_* for single-item retrieval. The only mild irregularity is get_live_figures not pairing with a list_/find_ family, but it still reads as the same grammar.
Seven tools is well within the ideal range and each one earns its place, forming a neat 3+3 matrix over two content types plus one live-data fetcher. There is no redundant or trivial tool in the set.
Discovery, ranked search, and whole-or-section retrieval are covered for both references and playbooks, with error payloads (suggestions, valid_values, available_sections) acting as recovery paths. Minor gaps: no dedicated tool for enumerating valid facet values or the list of guides with live figures, so an agent must trigger an error to learn them, and get_live_figures covers only three guides.