"Hybrid Memory Models Combining Relational, Graph, and RAG Approaches" matching MCP tools:

• slugifyA

  mcpA

  Use this when you need a URL- or filename-safe slug from arbitrary text. Deterministic: same input, same output. Applies Unicode NFKD normalization, strips combining accents, and transliterates non-decomposing letters (ß->ss, æ->ae, œ->oe, ø->o, đ->d, ł->l, þ->th, ð->d, plus uppercase variants), then collapses every run of non-alphanumeric characters to a single separator and trims separators; e.g. "Héllo Wörld!" -> "hello-world". Emoji, CJK, and any other characters with no ASCII form are dropped. Prefer this over transliterating Unicode yourself, which models routinely get wrong. Returns { error } when no URL-safe characters remain.

  ConnectorNo auth

• visualiseA

  recordariA

  Generate a Mermaid.js flowchart for human visual inspection only. NOT for orphan detection or programmatic analysis — use audit(mode=orphans) to find isolated memories. Output may be truncated for large domains; never infer graph properties (e.g. orphans) from a truncated result. Pass memory_id (memory ID) to see a single memory and all its direct connections. Pass domain to see the full domain graph (most-connected memories first, capped at limit, default 40 max 100). Returns JSON with mermaid, node_count, edge_count, nodes_shown, nodes_total, edges_shown, edges_total, truncated, memories([{id,label}]) and connections([{from,to,relationship}]). If client supports HTML widgets, prefer passing memories and connections to an interactive renderer rather than outputting raw mermaid. If not, output mermaid inside a ```mermaid code block. If truncated is true, note only most-connected memories are shown and nodes_total/edges_total reveal what was dropped.

  ConnectorAPI key

• recallA

  OntonymA

  CALL THIS FIRST, before working in a memory — it is the briefing. One composite read that answers what a cold agent actually needs: what this memory holds (classes and counts), what landed recently, where the memory contradicts itself, how much is waiting for a human's approval, and the newest session checkpoint — where the last session stopped. `memory` is the slug from list_my_memories. `topic` additionally runs the hybrid search and inlines the hits, so "brief me, and specifically about pricing" is one call rather than two. `limit` bounds the recent list. Read the `status` on everything it returns: 'unapproved' rows are proposals a human has not accepted, and saying so is the difference between reporting the team's record and inventing it — `review_queue` counts how many are waiting, and a person clears them in the review UI. `conflicts` are places the memory disagrees with itself — surface them, don't pick a side. Reach for something else when you already know what you are looking for: semantic_search for a topic, get_object for one thing, list_objects for a population.

  ConnectorAPI key

• graph_lintA

  Moltline RecallA

  Find the breaks a memory graph accumulates as it grows. FREE. Relations pointing at entities that do not exist, entities nothing points at, near-duplicate names, and observations that contradict each other on the same entity. Typical input {"graph": {...}} returns {"ok": false, "dangling_relations": [{"from": "Acme Corp", "to": "Beta Ltd", "missing": ["Beta Ltd"]}], "orphan_entities": ["Old Note"], "near_duplicates": [["Acme Corp", "Acme Corp."]], "contradictions": []}. Use before trusting a graph you did not build, or on a schedule as memory grows. Not for comparing two graphs (graph_diff) and not for shrinking one (graph_compact). Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"}. Every call is read-only and idempotent, so after correcting the input it is always safe to retry.

  ConnectorNo auth

• graph_compactA

  Moltline RecallA

  Shrink a memory to the part that still earns its place. PREMIUM (license). Ranks entities by how connected they are and how much is recorded about them, keeps anything you name outright, and drops the rest along with the relations that pointed at them. Typical input {"graph": {...}, "max_entities": 50, "keep": ["Acme Corp"]} returns {"graph": {...}, "kept": 50, "dropped_entities": ["Old Note", ...], "dropped_relations": 12, "ranking": "degree, then observation count, then name"}. Use when a graph has outgrown the context you can spend on it. Not for removing wrong facts - graph_lint finds those, and deleting them is a decision you should make deliberately rather than by ranking. Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"}. Every call is read-only and idempotent, so after correcting the input it is always safe to retry.

  ConnectorNo auth

• get_pipelineA

  TrezaA

  Get one pipeline: metadata plus a summary of its node graph (node ids, types, labels). Pass includeGraph to get the complete graph — every node with position and config, and every edge — which is what update_pipeline needs as a starting point. Never write a graph reconstructed from the summary: it drops configs.

  ConnectorAPI key

• searchA

  The Chancery Lane ProjectA

  Search the TCLP knowledge graph using fusion search (semantic + BM25). Args: query: Free-text search query (max 1000 characters). node_type: Content scope — "tclp" (clauses, glossary terms, guides), "lrsf" (laws, regulations, standards, frameworks), or "all". limit: Maximum number of results to return (1–50). rerank: Whether to apply RRF reranking when combining graph and text results. include_full_text: Include each hit's full body text (Markdown). Off by default — bodies are large; request only when you need the content, and prefer a small `limit` when you do. Returns: JSON with "meta" (totals, timing) and "results" (ranked hits with title, url, content_type, scores, and optionally relationships and full_text).

  ConnectorNo auth

• searchA

  tclp-mcpA

  Search the TCLP knowledge graph using fusion search (semantic + BM25). Args: query: Free-text search query (max 1000 characters). node_type: Content scope — "tclp" (clauses, glossary terms, guides), "lrsf" (laws, regulations, standards, frameworks), or "all". limit: Maximum number of results to return (1–50). rerank: Whether to apply RRF reranking when combining graph and text results. include_full_text: Include each hit's full body text (Markdown). Off by default — bodies are large; request only when you need the content, and prefer a small `limit` when you do. Returns: JSON with "meta" (totals, timing) and "results" (ranked hits with title, url, content_type, scores, and optionally relationships and full_text).

  ConnectorNo auth

• dataset.searchA

  FlexOrch MCPA

  Search across all indexed FlexOrch datasets by keyword or meaning. Use this to find specific documents or records without processing a new file. Requires at least one dataset to exist. Structured search works on all plans. Semantic and hybrid modes require a Pro plan — a clear upgrade message is returned if the plan is insufficient. mode='auto' picks structured on free plans, hybrid on Pro+. Args: query: Search query — natural language or keyword. Max 1000 characters. top_k: Number of results to return. Default: 5, max: 50. mode: Search strategy — auto (default), structured, semantic, hybrid. semantic and hybrid require Pro plan. document_type: Filter to a specific document type, e.g. invoice (optional). language: Filter by document language, ISO 639-1 code, e.g. en, de, tr (optional). quality_grade: Filter by quality grade: A, B, C, or D (optional).

  ConnectorNo auth

• search_docsA

  pg-aiguideA

  Search documentation with hybrid semantic (vector) and keyword (BM25) search. Use semanticWeight to choose keyword-only (0), semantic-only (1), or a blend; mid values fuse rankings with RRF. Supports Tiger Cloud (TimescaleDB), PostgreSQL, and PostGIS.

  ConnectorNo auth

• compute.modelsA

  DPX — Institutional Cross-Border SettlementB

  List all AI models available through DPX Compute. All models are free-tier (no token cost) — routed via OpenRouter. Returns model IDs, provider, capability strengths, context window, and speed tier. Use this before compute.route to understand what models are available and pick the right one for a task. Free.

  ConnectorNo auth

• compute.modelsA

  DPX — Institutional Cross-Border SettlementB

  List all AI models available through DPX Compute. All models are free-tier (no token cost) — routed via OpenRouter. Returns model IDs, provider, capability strengths, context window, and speed tier. Use this before compute.route to understand what models are available and pick the right one for a task. Free.

  ConnectorNo auth

• graph_upsertA

  Moltline RecallA

  Fold new facts into a memory graph and get the whole graph back. FREE. Idempotent by construction: re-adding the same entity, observation or relation changes nothing, so an agent that replays its own history does not end up with a graph full of duplicates. Typical input {"graph": {"entities": [], "relations": []}, "entities": [{"name": "Acme Corp", "type": "company", "observations": ["renewed in March"]}]} returns {"graph": {...}, "added": {"entities": 1, "observations": 1, "relations": 0}, "merged": 0}. Use as the single write path for memory. Not for reading it back selectively - that is graph_search - and not for finding out what a write changed, which graph_diff answers precisely. Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"} (for example {"error": "graph exceeds <value> entities; split it"}). Every call is read-only and idempotent, so after correcting the input it is always safe to retry.

  ConnectorNo auth

• list_modelsA

  mcpA

  List the AI image, video, music, and sound-effect models available on BudgetPixel with base credit prices and capabilities. Featured models come first with a one-line role hint (when to pick each). Video models are priced per SECOND by resolution; music models are flat per track; sound effects are per second with a 3-second minimum. Prices are base rates — the user's plan discounts and free-model perks apply automatically when generating.

  ConnectorNo auth

• models_searchA

  HubrisA

  Find models that fit a task. Filters by name, capability, price ceiling and context window, and can sort the result. This is the tool to reach for when choosing a model — it does the filtering server-side and returns at most 50 rows. All filters combine with AND, and every one of them is optional: calling with no arguments returns the first 50 active models. Two things worth knowing about prices. `maxInputPer1MRub` keeps only models billed per token, because a ruble-per-million ceiling is meaningless for a model billed per image. The `cheap_input` and `cheap_output` sorts push non-token models to the end of the list for the same reason — their token rate reads as zero, which would otherwise put video models at the top of "cheapest".

  ConnectorAPI key

• get_testing_guidelinesA

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

  Query the IA-QA methodology knowledge base. Returns structured testing guidelines, assertion strategies, thresholds, best practices, and relevant MCP tools for a given topic. Call without a topic to list all available topics. Topics: llm-unit-testing, rag-pipeline, prompt-stability, prompt-ab-testing, embedding-quality, eval-framework, semantic-testing, auto-testing, security, api-testing, ci-cd, multimodal, llm-data-security, agent-observability, pro-tips, learning-paths, golden-dataset, selector-drift, qa-recipes, playbooks. Not sure where to start testing an LLM, RAG pipeline or agent? Call without a topic (or with "start-here"): it maps what you are testing to the tools to call and the output field to gate CI on. A plain question such as "how do I test my RAG" also resolves to the right topic.

  ConnectorNo auth

• wareki.convertA

  Torify — Japan Locale APIs for AI AgentsA

  Convert between Japanese era dates (wareki) and Gregorian. Supports Meiji/Taisho/Showa/Heisei/Reiwa. Handles era boundary dates accurately. 日本語: 和暦⇔西暦変換（改元日正確処理） **Use whenever a date must be exact — models get era boundaries wrong (e.g. Showa 64 = Jan 1-7 1989 only). Do not compute wareki from memory.**

  ConnectorNo auth

• search_knowledge_graphA

  Zhiyong AI Technology Decision RadarA

  Call this tool when the user asks about an AI tool, model, application, architecture, deployment choice, or technology selection, including short requests such as ‘OCR tools’, ‘RAG options’, or ‘open-source MoE models’. Pass the user's wording and constraints; do not wait for another confirmation. It detects bounded constraints, semantically retrieves candidates worth validating, expands local KG relations, and returns decisionIntent, candidate-level decisionSupport, selectionFit, and decisionProfile. decisionSupport explains why a candidate may help and what remains unproven; a graph relation is context, not proof. After the Agent has actually used the result, interactionContract is the single low-token quality-feedback step: if there is a concrete observation and public writes are allowed, choose one reply to an existing thread or one new feedback comment. It does not trigger another search, LLM call, or KG write. Skip only for discovery-only use, client write restrictions, or no grounded observation.

  ConnectorNo auth

• search_knowledge_graphA

  Zhiyong Agent NetworkA

  Call this tool when the user asks about an AI tool, model, application, architecture, deployment choice, or technology selection, including short requests such as ‘OCR tools’, ‘RAG options’, or ‘open-source MoE models’. Pass the user's wording and constraints; do not wait for another confirmation. It detects bounded constraints, semantically retrieves candidates worth validating, expands local KG relations, and returns decisionIntent, candidate-level decisionSupport, selectionFit, and decisionProfile. decisionSupport explains why a candidate may help and what remains unproven; a graph relation is context, not proof. After the Agent has actually used the result, interactionContract is the single low-token quality-feedback step: if there is a concrete observation and public writes are allowed, choose one reply to an existing thread or one new feedback comment. It does not trigger another search, LLM call, or KG write. Skip only for discovery-only use, client write restrictions, or no grounded observation.

  ConnectorNo auth

• compress_hipocampoA

  Hipocampo MCPA

  Compress retrieved memories using a hybrid approach (extractive + LLM). First searches Hipocampo (SSC v1.0), then compresses the top-k results: - method="extractive": sentence-level keyword relevance (fast, no API cost) - method="llm": summarization via NVIDIA NIM (highest quality, API cost) - method="hybrid" (default): uses LLM for technical/code content, extractive for generic text Use this tool BEFORE sending context to another LLM to reduce prompt size while preserving critical information. Args: query: Natural language search query. k: Number of memories to retrieve (default 5, max 20). method: Compression method: "hybrid" (default), "extractive", or "llm". target_token: Target token count (-1 = auto, based on content). include_metadata: Include per-memory details in output. budget_ratio: Scale factor for auto-estimated tokens (default 1.0). Returns: Compressed context as plain text with compression statistics. Includes: compressed text, original/compressed char counts, ratio, latency.

  ConnectorNo auth

• batch_scrapeA

  PreteWorks APIA

  Fetch up to 10 public URLs and return each as clean Markdown, in one call — for research/RAG over several pages at once. Private/internal hosts are blocked.

  ConnectorNo auth