Deterministic knowledge for AI agents: search, fetch, cite and verify a 48M-entry knowledge base.
Scored across 16 tools
Most tools have distinct targets: search vs search_global vs batch_search differ by scope and snapshot consistency, while resolve/fetch/cite/verify form a clear retrieval-and-verification lifecycle. Minor overlap remains among search, browse_documents, and stats-with-query, but descriptions should usually prevent misselection.
All tool names use the coldstate_ prefix and snake_case consistently, with predictable verb/noun or noun forms. No camelCase or competing conventions appear, and variants like search_global and batch_search fit the same pattern.
16 tools is slightly above the ideal range, but each maps to a distinct knowledge-infrastructure capability such as search, retrieval, citation, verification, or metadata. The set is not bloated enough to be confusing, though a few metadata tools could potentially be consolidated.
The core lifecycle—search, resolve, fetch, cite, verify, related/isomorph, and stats—is well covered for a deterministic knowledge server. Gaps include no visible index creation, ingestion, or update tools, so users cannot manage knowledge sources fully through this surface.
Run several knowledge-base queries in one call against a single consistent kb_snapshot. Returns each query with its ranked results — efficient and snapshot-consistent across the whole batch.
Inputs: queries (required), limit, domain.
Browse documents in a ColdState index. Indexes are zero-knowledge, so each document is returned as your ref_id plus its state classification and E-score — no stored text. Map ref_id back to your own content.
Inputs: index_id (required), limit, offset.
Machine-readable manifest of this server: available tools, knowledge domains, current kb_snapshot, determinism guarantees, and limits. Call this first to self-configure.
Inputs: no named parameters.
Get a canonical, reproducible citation for a knowledge entry by id (title, domain, source, content_hash, kb_snapshot). Cite a fact so it can be re-verified later with coldstate_verify.
Inputs: id (required).
List all available knowledge domains in ColdState's global knowledge base with entry counts.
Inputs: no named parameters.
Explain why a specific document ranked for a query. Returns a per-term relevance breakdown and an overall match summary — deterministic and reproducible.
Inputs: query (required), doc_id (required), index_id (required).
Fetch a single knowledge-base entry verbatim by its stable id (e.g. "eng_000229f438f7cebd"). Deterministic retrieval — returns the full record plus a content_hash and kb_snapshot for reproducible citation.
Inputs: id (required).
Cross-domain structural analog: given a knowledge entry id, find its closest structural twin in each OTHER domain ("same structure, different domain"). Deterministic. Use coldstate_resolve to turn a name into an id first.
Inputs: id (required), limit.
List all your ColdState indexes with their status, mode, document count, and domain preset.
Inputs: no named parameters.
Resolve a name, alias, or surface form to its canonical ColdState knowledge entry (id + title). Deterministic: an exact title match wins, else the top relevance-Ψ entry. Returns the canonical entry plus alternatives. Use the returned id with coldstate_fetch / coldstate_related.
Inputs: query (required), limit.
Search a ColdState index by collection name or index ID. Returns ranked results with Ψ scores and state classification (CRYSTALLINE/FLUID/REACTIVE/DECOHERENT). Provide exactly one of collection or index_id.
Inputs: query (required), limit, offset, index_id, +1 more.
Search ColdState's global knowledge base (48.4M+ entries across 35 domains including SCIENCE, MEDICINE, TECHNOLOGY, HISTORY, etc). Returns deterministically ranked results with Ψ relevance scores and a state classification. Optionally filter by domain.
Inputs: query (required), limit, domain, offset.
Get the current knowledge-base snapshot/version id. Pin this kb_snapshot for reproducibility: the same query against the same kb_snapshot returns identical results.
Inputs: no named parameters.
Deterministic statistics. With a query: total matches + state distribution of the top results. Without: global knowledge-base stats (entry count, domains, snapshot).
Inputs: query, domain.
Verify a previously-cited knowledge fact is unchanged. Provide the entry id and the content_hash you stored earlier; returns whether it still matches the current knowledge base. The trust primitive for AI-to-AI fact-checking.
Inputs: id (required), content_hash (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Free verified network knowledge for AI agents: deterministic answers, honest unknowns.
Deterministic fact verification for AI agents — checksums & curated data, not guesses.
Open scientific and engineering knowledge for AI agents: search, evidence, document publishing.
Persistent working memory for AI agents: a knowledge graph with semantic search and decay.
Add one secure layer between your agents and this server.
