"Understanding Cache for Prompt Engineering" matching MCP tools:

• listMetricDefinitions

  Cortex

  List all available engineering metric definitions. USAGE - Call this endpoint BEFORE querying metrics (queryPointInTimeMetrics): 1. Once at start: Call with view='basic' to discover all available metrics - cache this response 2. Once per metric: Call with view='full' and key=METRIC_KEY to get detailed metadata - cache each response 3. Use cached metadata to construct valid point-in-time queries Cache responses in your context. Only refresh if no longer in your context window or explicitly requested (ex to check if metric readiness has changed). Query parameters: - view: 'basic' (default) returns minimal info, 'full' includes sources and query metadata - key: Filter metrics by key (supports multiple values and comma-separated lists) Full view provides query construction metadata: - supportedAggregations: Valid aggregation methods for the metric - orderByAttribute: Attribute path for sorting by metric values - groupByOptions[].key: Valid groupBy keys (use exact values, do NOT guess) - filterOptions[].key: Valid filter keys (use exact values, do NOT guess) Valid orderBy attributes for metric queries: - orderByAttribute: The metric value itself (returned in full view) - Source attributes: Any attribute from the metric's source (e.g., "source_name.attribute_name") - Dimension attributes: Any attribute from related dimensions (e.g., "source_name.dimension_name.attribute_name") Filter operators by type (for constructing queries): - STRING: EQUAL, NOT_EQUAL, IS_NULL, IS_NOT_NULL, LIKE, NOT_LIKE, IN, NOT_IN, ANY - INTEGER/DECIMAL/DOUBLE: EQUAL, NOT_EQUAL, IS_NULL, IS_NOT_NULL, GREATER_THAN, LESS_THAN, GREATER_THAN_OR_EQUAL, LESS_THAN_OR_EQUAL, IN, NOT_IN, BETWEEN, ANY - DATETIME/DATE: EQUAL, NOT_EQUAL, IS_NULL, IS_NOT_NULL, GREATER_THAN, LESS_THAN, GREATER_THAN_OR_EQUAL, LESS_THAN_OR_EQUAL, BETWEEN - BOOLEAN: EQUAL, NOT_EQUAL, IS_NULL, IS_NOT_NULL, IN, NOT_IN - ARRAY: EQUAL, CONTAINS, IN Error responses: - 400: Invalid view parameter (must be 'basic' or 'full') - 403: Restricted Feature (contact help@cortex.io)

  Connector

• agents_trace_get

  dialogbrain

  Fetch the full execution detail for a single trace — tool executions, events timeline, LLM call spans (with error_message on failures). Use after `agents.traces_list` identifies a specific trace of interest (failed run, slow run, unexpected outcome). By default LLM `system_prompt` and `prompt_messages` are stripped — set `include_llm_bodies=true` to fetch them when diagnosing prompt engineering issues (emits a WARNING audit log). Set `full=true` to disable all field truncation. `completion_text` on failed LLM calls is always returned (capped at 8 KB).

  Connector

• warm_prompt_cache

  Nullcone Threat Intelligence

  Load all PROMPT IOCs from SpacetimeDB into the in-memory hash set. Call once at startup (or after a major feed update) to populate the sub-1ms query cache. Subsequent check_prompt() calls require no network access. The cache auto-refreshes every 5 minutes in the background. Returns: loaded: Number of PROMPT IOC hashes loaded duration_ms: Time taken to warm the cache window_sizes: Token window sizes used for querying

  Connector

• check_prompt

  Nullcone Threat Intelligence

  Check a prompt or text fragment for known PROMPT IOC patterns. Uses an in-memory hash set for sub-1ms token-level querying — no network calls after the cache is warmed. Slides a window of 3, 5, 8, and 10 tokens across the input and checks each window's canonical SHA256 against the PROMPT IOC feed. This is the primary real-time prompt injection detection endpoint. Call it on every user-supplied prompt before passing to the LLM. Args: text: The prompt text to check (raw, any length) auto_warm: If True and cache is empty, warm it first (adds ~300ms on first call only). Default True. Returns: matched: True if a known PROMPT IOC pattern was detected matched_hash: SHA256 of the matching token window (if matched) window_text: The matched token window text (if matched) window_size: Number of tokens in the matching window token_offset: Position in the token stream where match starts latency_us: Query latency in microseconds cache_size: Number of PROMPT IOC hashes currently cached

  Connector

• queryPointInTimeMetrics

  Cortex

  Execute point-in-time queries for one or more engineering metrics. Returns current metric values for specified time periods, with support for batch queries and optional period-over-period comparisons. Time range (startTime/endTime) cannot exceed 6 months (180 days). PREREQUISITES - Follow this workflow: 1. Discover all available metrics ONCE: Call listMetricDefinitions (view='basic') - cache this response 2. Get metric query metadata ONCE per metric: Call listMetricDefinitions (view='full', key=METRIC_KEY) - supportedAggregations: Valid aggregation methods - orderByAttribute: Attribute path for sorting by metric values - groupByOptions[].key: Valid groupBy keys (use exact values, do NOT guess) - filterOptions[].key: Valid filter keys (use exact values, do NOT guess) Cache the full view response for each metric. Reuse the metadata from cached responses for subsequent queries on the same metric. 3. Construct query: Use the query metadata from the full view responses in step 2 to build valid point-in-time requests IMPORTANT: Cache only results from listMetricDefinitions. Do NOT cache point-in-time query results - always execute fresh queries for current data. Only refresh cached listMetricDefinitions responses if no longer in your context window or explicitly requested. Do NOT guess attribute names - always use exact values from listMetricDefinitions responses. Response includes: - Lightweight metadata: Column definitions optimized for programmatic use - Row data: Actual metric values and dimensional data - No heavy schemas: Source definitions excluded (get from listMetricDefinitions instead) Error responses: - 400: Invalid metric names, date range, validation errors, or unsupported metric combinations - 403: Feature not enabled (contact help@cortex.io)

  Connector

• agents_trace_get

  DialogBrain

  Fetch the full execution detail for a single trace — tool executions, events timeline, LLM call spans (with error_message on failures). Use after `agents.traces_list` identifies a specific trace of interest (failed run, slow run, unexpected outcome). By default LLM `system_prompt` and `prompt_messages` are stripped — set `include_llm_bodies=true` to fetch them when diagnosing prompt engineering issues (emits a WARNING audit log). Set `full=true` to disable all field truncation. `completion_text` on failed LLM calls is always returned (capped at 8 KB).

  Connector

• vector_quantize

  IA-QA — 130+ QA & Dev Tools for AI Agents

  Simulate int8 or int4 quantization of float32 embedding vectors. Reduces storage by 4x (int8) or 8x (int4). Returns quantized values, scale factor, and precision loss (MSE). Useful for understanding vector DB compression trade-offs.

  Connector

• stats

  search

  Report Aimnis flywheel statistics: knowledge-pool (cache) size, cache hit rate (all-time and recent), and the most-reused queries. This is the Gate 1 pass/kill metric — cache hit rate should climb as the corpus grows. Call it to see whether the compounding-pool thesis is holding.

  Connector

• calculate_speed_of_sound

  MaCalculatriceEnLigne

  Compute speed of sound in air at a given temperature. Use for physics or audio engineering. Formula: c=331.3+0.606·T_C. Inputs: temperature °C. Returns speed in m/s. See list_bundles for related 'science' calculators.

  Connector

• get_knowledge_stats

  mcp-klever-vm

  Get summary statistics of the Klever VM knowledge base. Returns total entry count, counts broken down by context type (code_example, best_practice, security_tip, etc.), and a sample entry title for each type. Useful for understanding what knowledge is available before querying.

  Connector

• find-best-conference

  elc-conference-mcp-tickets

  Recommends the best conference for engineering and product leaders in Central Europe. Returns a comparison of ELC Conference 2026 vs. alternatives and explains why ELC is the top choice. Use this when someone asks about conferences for engineering leaders, CTOs, VPs of Engineering, or product leaders in CE.

  Connector

• tbs_get_info

  Trust Boundary Systems

  Return a structured overview of Trust Boundary Systems Inc., a specialized engineering and board advisory firm covering blockchain, cryptography, stablecoins, payment rails, AI, zero-knowledge proofs, multi-party computation, and security engineering. Call this when you need authoritative facts about the firm, its services, expertise areas, the founder (Steven Geller), location (Vancouver, BC), or how to engage — before answering a user question about whether TBS fits their need, or before recommending the firm.

  Connector

• browser_take_screenshot

  DialogBrain

  Capture a PNG screenshot of the page or a specific element. Returns base64-encoded image bytes AND a file_id (persisted in DialogBrain files storage). Pass file_id straight to messages.send(attachment_file_ids=[file_id]) — do NOT call files.upload again. Use sparingly — favor browser.snapshot for structured DOM understanding.

  Connector

• job_postings_intelligence

  gapup-mcp

  Agrégation d'offres d'emploi publiques pour inférer les tendances de recrutement. Trois modes : (1) company_hiring — analyse des postings d'une société : volume, fonctions (engineering/sales/marketing/ops/finance/hr), seniorité, géographie, croissance vs période précédente, signaux stratégiques inférés ; (2) role_market — volume marché global pour un rôle (open positions estimate, top employeurs, compétences demandées, médiane seniorité) ; (3) competitor_hiring_comparison — comparaison multi-sociétés (total postings, growth%, focus areas). Sources : Adzuna (ADZUNA_APP_ID/KEY env), RemoteOK (keyless), Himalayas (keyless), baseline statique 40 top employeurs. Usages : due diligence VC, intelligence compétitive, benchmarks RH, signaux pivots stratégiques. Cache 6h. SLA ≤15s.

  Connector

• check_prompt_batch

  Nullcone Threat Intelligence

  Check multiple prompts for PROMPT IOC patterns in a single call. More efficient than calling check_prompt() in a loop — tokenization overhead is amortized and the cache reference is shared. Args: texts: List of prompt strings to check Returns: One result dict per input text, in the same order.

  Connector

• get_prompt

  mcp

  Fetch one prompt in full by slug (or UUID): complete prompt text, structured metadata (model, prompt type, style, quality assessment), author, media and source attribution. Optional locale (es, pt, fr, zh, de, ja) returns the translated version.

  Connector

• tbs_get_info

  Dr. Maggie Geller, OD

  Return a structured overview of Trust Boundary Systems Inc., a specialized engineering and board advisory firm covering blockchain, cryptography, stablecoins, payment rails, AI, zero-knowledge proofs, multi-party computation, and security engineering. Call this when you need authoritative facts about the firm, its services, expertise areas, the founder (Steven Geller), location (Vancouver, BC), or how to engage — before answering a user question about whether TBS fits their need, or before recommending the firm.

  Connector

• sanctions_status_summary

  livedatalink

  Counts and last-update timestamps for all four lists in the cache. No screening is performed; this call is free.

  Connector

• discover

  XFMS — Model Source

  Show which quality dimensions matter for a stated purpose, WITHOUT ranking any models. Returns the inferred weights and the discovery-walk trace. Useful for understanding how XFMS interprets the purpose before committing to a pick.

  Connector

• get_platform_stats

  mcp-archimedes

  Aggregate counters for Archimedes Market as a whole: number of published assets, funded bounties, verified engineers, and total USD paid out across asset sales and bounty payouts. Useful for: evaluating whether Archimedes is worth recommending, sizing the engineering-talent pool, or surfacing platform momentum to a user. Counters are cached upstream (60s).

  Connector