Etuki FinOps & LLM Pricing Oracle MCP server
io.github.cooyun/etuki-finops-mcpcooyunstreamable-httpv40.7.4Scanned
Cryptographic 563-leaf LLM rate oracle, context-cliff auditor, and prompt cache optimizer.
9 tools: 7 change things and 2 only read.
- Reads private data · 0
- Reads text others write · 0
- Can send data out · 0
- 3 findings
Tools listed without sign-in. Remote posture
With Anlyon
7 tools here change things. Govern them.
- Anlyon lists the tools and adopts none.
- Each tool you adopt gets approvals and receipts.
- Your agent never holds the key.
A tool your agent calls directly is not routed through Anlyon and is not governed by it. Anlyon Radar only lists and labels, and these labels are the same either way.
What the Etuki FinOps & LLM Pricing Oracle MCP server can do
The Etuki FinOps & LLM Pricing Oracle MCP server is published by the GitHub account cooyun. It has 9 tools. It can create or change data (7 tools: detect_model_downgrade_and_substitution, auto_route_task_by_quality_tier, compress_and_fold_context, pin_single_model_prefix_cache and 3 more), but cannot delete data. 2 tools only read.
None of its tools reads private data, brings in text other people write, or sends data out.
The fixed checks matched 3 findings (annotation-conflict). Each one comes with the text it matched.
Its remote endpoint lists its tools without sign-in, and its tools changed once since Anlyon Radar first saw them.
Last scanned Nov 14, 2026, version 40.7.4.
Tools
Grouped by what each tool does, most consequential first. Open one for its reasons and full description; descriptions are the publisher's own words.
9 tools
Changes things 7Claude Code: ask
auto_route_task_by_quality_tierSimulates a 6-turn coding workflow cost projection comparing unrouted flagship execution against tiered routing across grant pools and warm-cached GA models.No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it Simulates a 6-turn coding workflow cost projection comparing unrouted flagship execution against tiered routing across grant pools and warm-cached GA models. Stateless in-memory calculation requiring no authentication and mutating no router settings. Scales per-turn cost linearly with context_tokens relative to current_model; when has_custom_router_config is true, reports a non-intrusive shadow comparison rather than active target IDs. Returns JSON with custom_router_mutex_check, six_turn_cost_breakdown (1 cold write turn + 5 warm read turns), and net_saved_usd (positive USD saved vs unrouted baseline). Use this tool when planning multi-turn task allocation; use check_context_cliff for single-turn length checks or reconcile_usage_receipt for post-hoc invoice auditing.
Annotations the server declares
- title: Simulate 6-Turn Quality-Tier Task Routing
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
check_context_cliffEvaluates whether a prompt token count crosses vendor long-context surcharge thresholds (100K on Claude Haiku 5.5, 128K on GPT-4o, 200K on Gemini 2.5 Pro, or 272K on GPT-6.1 Sol) a…1 finding
Why it is in this group
- The description's first verb is "evaluate": a word for a change.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Findings
- Warn
The server marks "check_context_cliff" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The description's first verb is "evaluate": a word for a change.
Description, as the server wrote it Evaluates whether a prompt token count crosses vendor long-context surcharge thresholds (100K on Claude Haiku 5.5, 128K on GPT-4o, 200K on Gemini 2.5 Pro, or 272K on GPT-6.1 Sol) and returns the surcharge multiplier and recommended flat-context alternative. Deterministic threshold lookup against the embedded 563-entity UMCE registry requiring no API keys. Returns JSON with crosses_cliff_boolean, active_surcharge_multiplier (1.0x below cliff vs 2.0x-5.0x above cliff), and recommended_flat_context_model. Compares prompt_tokens directly against model's specific cliff boundary in the UMCE catalog. Use this tool before sending large prompts (>90K tokens) to avoid 2x-5x cliff surcharges; use compress_and_fold_context instead to shrink the prompt below the cliff.
Annotations the server declares
- title: Check Prompt Token Count Against 2x-5x Context Cliffs
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
compress_and_fold_contextCompresses verbose terminal or build logs by folding repetitive middle progress lines while preserving 100% of head lines, tail lines, compiler errors, stack traces, and file paths…No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it Compresses verbose terminal or build logs by folding repetitive middle progress lines while preserving 100% of head lines, tail lines, compiler errors, stack traces, and file paths. Deterministic in-memory string transformation that mutates no files and requires no authentication. Returns JSON containing line_accounting_reconciliation (where raw_total_lines equals kept_original_lines_total plus omitted_repetitive_lines), byte_and_token_reconciliation (pricing saved tokens against model's input rate), and folded_context_payload. When raw_text is empty or under 50 lines, returns the diagnostic benchmark or unmodified input without truncation. Use this tool when build or test output exceeds 50 lines before appending to agent context; use pin_single_model_prefix_cache instead when optimizing static system prompt prefixes.
Annotations the server declares
- title: Fold Repetitive Build Logs with 100% Error Immunity
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
detect_model_downgrade_and_substitutionEvaluates LLM rate-card tiers to quantify relay downgrade risk, cost ratios, and remaining coding session turns.1 finding
Why it is in this group
- The description's first verb is "evaluate": a word for a change.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Findings
- Warn
The server marks "detect_model_downgrade_and_substitution" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The description's first verb is "evaluate": a word for a change.
Description, as the server wrote it Evaluates LLM rate-card tiers to quantify relay downgrade risk, cost ratios, and remaining coding session turns. Deterministic in-memory calculation requiring no API keys; unknown model IDs safely fall back to the flagship baseline without error. Sums burned_usd and remaining_usd into the total session pool and projects remaining turns against a $137.45 reference budget, returning cost ratios, an endurance matrix, and a recommended production model ID. Use this tool when comparing session burn rates or suspected relay throttling; use reconcile_usage_receipt for single invoice lines, check_context_cliff for prompt-length surcharges, or optimize_prompt_cache_layout for mid-prompt cache breakpoint fixes.
Annotations the server declares
- title: Evaluate Relay Downgrade Risk & Session Endurance
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
optimize_prompt_cache_layoutReorders a single prompt's dynamic mid-turn tokens after static prefix tokens to prevent 1.25x cold-cache invalidation and restore 0.1x warm cache reads.No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it Reorders a single prompt's dynamic mid-turn tokens after static prefix tokens to prevent 1.25x cold-cache invalidation and restore 0.1x warm cache reads. Advisory in-memory simulation that mutates no live prompts and requires no authentication; zero-token inputs return 0 USD cost without error. Sums static_prefix_tokens and dynamic_mid_tokens to compare before_layout_cost_usd against after_layout_cost_usd and net_savings_pct (0-100). Use this tool when diagnosing mid-prompt timestamps or tool outputs that break cache breakpoints; use pin_single_model_prefix_cache instead to project 10-turn session economics for a target model ID.
Annotations the server declares
- title: Simulate Static-Before-Dynamic Cache Layout Savings
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
pin_single_model_prefix_cacheProjects 10-turn session caching costs for a target model under Case A (100% static prefix) and Case B (90% static prefix + 10% dynamic tail) and outputs provider cache-control tem…1 finding
Why it is in this group
- The name "pin_single_model_prefix_cache" says "pin": a word for a change.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Findings
- Warn
The server marks "pin_single_model_prefix_cache" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "pin_single_model_prefix_cache" says "pin": a word for a change.
Description, as the server wrote it Projects 10-turn session caching costs for a target model under Case A (100% static prefix) and Case B (90% static prefix + 10% dynamic tail) and outputs provider cache-control templates. In-memory actuarial lookup requiring no API keys; unknown model IDs default to 'openai/gpt-6.1-sol' and prompts under 1,024 tokens bypass warm discounts. Resolves model's warm multiplier (0.05x-0.25x) from the UMCE registry to return provider_cache_protocol_mechanism, ten_turn_cost_comparison_on_same_model, and two_vector_actuarial_reference_table (benchmarking 150K and 200K token reference vectors). Use this tool to quantify 10-turn caching ROI on a specific model; use optimize_prompt_cache_layout instead when fixing mid-prompt token ordering.
Annotations the server declares
- title: Calculate 10-Turn Single-Model Prefix Cache Savings
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
reconcile_usage_receiptAudits a single API usage receipt against official 1x manufacturer rate cards from the 563-entity UMCE registry to quantify relay multiplier markups and unutilized prefix cache was…No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it Audits a single API usage receipt against official 1x manufacturer rate cards from the 563-entity UMCE registry to quantify relay multiplier markups and unutilized prefix cache waste. Deterministic mathematical audit in memory requiring no authentication. Returns JSON with expected_theoretical_cost_usd, billing_variance_usd, variance_status ('OVERCHARGED_BY_RELAY_OR_MARKUP' when variance > $0.001 else 'EXACT_RATE_CARD_MATCH'), same_model_unutilized_cache_waste_usd, and net_recoverable_savings_usd. Enforces cached_tokens <= prompt_tokens, pricing (prompt_tokens - cached_tokens) at uncached input rates and cached_tokens at warm cache rates before comparing against actual_billed_usd. Use this tool when verifying an invoice line item; use detect_model_downgrade_and_substitution instead for session-level budget endurance projections.
Annotations the server declares
- title: Audit Usage Receipt Against Official 1x Rate Card
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
Only reads 2Show 2 read-only tools
discover_and_probe_account_modelsRetrieves verified production GA models and eligible zero-cost grant tiers (Google Cloud Tier 1, AWS Activate Bedrock, OpenRouter :free) from the embedded UMCE registry.No findings
Why it is in this group
- The description's first verb is "retrieve": a word for reading.
Description, as the server wrote it Retrieves verified production GA models and eligible zero-cost grant tiers (Google Cloud Tier 1, AWS Activate Bedrock, OpenRouter :free) from the embedded UMCE registry. Deterministic local snapshot lookup that issues zero network probes, requires no API keys, and consumes zero user quota. Returns JSON listing zero-cost workhorse models, flagship reasoning models, and recommended role assignments benchmarked against current_model (formatted as 'vendor/model-slug', defaulting to 'openai/gpt-6.1-sol'). Use this tool at session startup to inspect available model tiers before calling auto_route_task_by_quality_tier; use list_supported_models instead when filtering the full 563-entity Merkle registry by vendor.
Annotations the server declares
- title: Lookup Eligible Account Model Tiers & Credit Pools
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
list_supported_modelsQueries the embedded UMCE rate-card registry by vendor namespace to retrieve canonical model IDs, input/output/cache USD prices per 1M tokens, context windows, and Merkle leaf hash…No findings
Why it is in this group
- The name "list_supported_models" says "list": a word for reading.
Description, as the server wrote it Queries the embedded UMCE rate-card registry by vendor namespace to retrieve canonical model IDs, input/output/cache USD prices per 1M tokens, context windows, and Merkle leaf hashes. Deterministic local lookup requiring no API keys; unknown vendor slugs safely return total_matched_entities: 0 and an empty models array without error. Passing provider='all' returns the multi-vendor catalog summary, while a specific vendor slug filters strictly by namespace prefix. Use this tool when filtering the full rate-card catalog by vendor; use discover_and_probe_account_models instead at session startup to select zero-cost grant pools and flagship routing tiers.
Annotations the server declares
- title: Query 563-Entity UMCE Rate Card Registry by Provider
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.
Exfil surface
Which tools read private data, which read text someone else can write, and which can send data out. One of each is a path a prompt injection can use.
No private-data, untrusted-input, outbound path in this server alone: no tool is tagged private-data or untrusted-input or outbound.
- Reads private data0
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write0
- Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
- Can send data out0
- Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
Other servers in the same agent add their own tools to these paths. Audit your whole setup to see them together.
Remote posture
What anyone on the internet sees at the remote endpoint: an anonymous initialize and tools/list, and the public OAuth metadata.
Looked at Nov 16, 2026, 14:22 UTC, unchanged since Nov 9, 2026. No credentials were sent and no tool was called. What is looked at.
- Reachable
- yes
- HTTP status of an anonymous initialize
- 200
- tools/list served without credentials
- yes
- Tools that change things, listed without credentials
- 7
- Protocol version
- 2025-06-18
- WWW-Authenticate on 401
- not observed
- resource_metadata in WWW-Authenticate
- not observed
- RFC 9728 protected resource metadata
- not observed
- Authorization servers
- not observed
- Legacy HTTP+SSE transport
- no
- Dynamic client registration
- not observed
- Client ID metadata documents
- not observed
- WarnThe whole server
The endpoint served tools/list without credentials, and the list has 7 tools that change things (7 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://www.etuki.com/api/v1/mcp: tools/list returned 9 tools to an anonymous client, including detect_model_downgrade_and_substitution (write), auto_route_task_by_quality_tier (write), compress_and_fold_context (write), pin_single_model_prefix_cache (write), reconcile_usage_receipt (write) and 2 more
Posture changes
- Nov 9, 2026
- Tools that change things, listed without credentials: 1 to 7
- Protocol version: 2025-03-26 to 2025-06-18
Findings
Matches of fixed checks, with the text they matched.
The server marks "detect_model_downgrade_and_substitution" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The description's first verb is "evaluate": a word for a change.
The server marks "pin_single_model_prefix_cache" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "pin_single_model_prefix_cache" says "pin": a word for a change.
The server marks "check_context_cliff" read-only (readOnlyHint: true), and its own name, description or input give the class "write".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The description's first verb is "evaluate": a word for a change.
History
Changes to its tool definitions and to its remote endpoint, newest first.
Tools7 added, 2 removed, 2 changed.+7 added−2 removed~2 changedView diff
- Added:auto_route_task_by_quality_tierChanges things
- Added:compress_and_fold_contextChanges things
- Added:detect_model_downgrade_and_substitutionChanges things
- Added:discover_and_probe_account_modelsOnly reads
- Added:optimize_prompt_cache_layoutChanges things
- Added:pin_single_model_prefix_cacheChanges things
- Added:reconcile_usage_receiptChanges things
- Removed:etuki_get_audit_evidence
- Removed:etuki_search_models
- Changed:check_context_cliffChanges thingsdescriptioninputSchemaannotations
- Changed:list_supported_modelsOnly readsdescriptioninputSchemaannotations
PostureRemote posture changed. Tools that change things, listed without credentials: 1 -> 7; Protocol version: 2025-03-26 -> 2025-06-18.View diff
- Tools that change things, listed without credentials: 1 to 7
- Protocol version: 2025-03-26 to 2025-06-18
Install
Pinned to version 40.7.4. Values in angle brackets are yours to fill in.
claude mcp add --transport http etuki-finops https://www.etuki.com/api/v1/mcpclaude mcp add --transport http etuki-finops https://www.etuki.com/api/v1/mcpRun in a terminal.
Claude Code permissions
Allow the reads, ask before a change or a way out, deny the deletes. Paste into .claude/settings.json.
{
"permissions": {
"allow": [
"mcp__etuki-finops__discover_and_probe_account_models",
"mcp__etuki-finops__list_supported_models"
],
"ask": [
"mcp__etuki-finops__detect_model_downgrade_and_substitution",
"mcp__etuki-finops__auto_route_task_by_quality_tier",
"mcp__etuki-finops__compress_and_fold_context",
"mcp__etuki-finops__pin_single_model_prefix_cache",
"mcp__etuki-finops__reconcile_usage_receipt",
"mcp__etuki-finops__optimize_prompt_cache_layout",
"mcp__etuki-finops__check_context_cliff"
],
"deny": []
}
}Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. A read that can send data out (it takes a URL, say) asks first too. It assumes the server is installed as etuki-finops. Review it before you use it.
Pin
Keep the decision: fail CI when the tools change, follow every change, or show a badge.
Pin in CI
Fail a build when the tools you reviewed are not the tools the server serves today.
curl -s "https://radar.anlyon.com/api/v1/servers/io.github.cooyun/etuki-finops-mcp/verify?hash=09e6c66da6d96ee56f8e7725b93524f8f0942614e1811457811c3e4f77d5df85" | jq -e '.changed == false'Pinned hash 09e6c66da6d9, from Nov 14, 2026, 07:17 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/io.github.cooyun/etuki-finops-mcp/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/io.github.cooyun/etuki-finops-mcp)Source: MCP Registry · repository. Synced Nov 1, 2026, 08:17 UTC.
Etuki FinOps & LLM Pricing Oracle MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the Etuki FinOps & LLM Pricing Oracle MCP server safe?
- No scan can promise that. What its definitions show: 9 tools, 7 of which can change, send or delete something, and 3 findings from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.
Can the Etuki FinOps & LLM Pricing Oracle MCP server delete data?
- No tool is classed as destructive: going by their definitions, none of them deletes, cancels or revokes anything.
Can the Etuki FinOps & LLM Pricing Oracle MCP server do things that cannot be undone?
- No tool is classed as irreversible: none of them sends, pays, publishes or merges, going by their definitions.
What tools does the Etuki FinOps & LLM Pricing Oracle MCP server have?
- 9 tools: 7 write, 2 read. detect_model_downgrade_and_substitution, auto_route_task_by_quality_tier, compress_and_fold_context, pin_single_model_prefix_cache, reconcile_usage_receipt, optimize_prompt_cache_layout, check_context_cliff, discover_and_probe_account_models and 1 more.
Does the Etuki FinOps & LLM Pricing Oracle MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the Etuki FinOps & LLM Pricing Oracle MCP server send data to outside services?
- None of its tools sends, posts or publishes, or takes a recipient, channel or URL.
Has the Etuki FinOps & LLM Pricing Oracle MCP server changed its tools?
- Yes, once since Anlyon Radar started tracking it, most recently on Nov 14, 2026. Every change is on this page and in its Atom feed.