# CoreModels (remote · go.coremodels.io)

Model, map and validate data schemas. Export JSON Schema, JSON-LD, LinkML, OWL, ShEx, SQL, Avro.

- Trust score: 72/100 (medium)
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-21

## Components

- remote · `go.coremodels.io`: 72/100 (this document), [markdown](https://verifymcp.io/servers/io-coremodels-coremodels/go.md), [page](https://verifymcp.io/servers/io-coremodels-coremodels/go)
- remote · `go.coremodels.io`: 36/100, [markdown](https://verifymcp.io/servers/io-coremodels-coremodels/mcp-admin.md), [page](https://verifymcp.io/servers/io-coremodels-coremodels/mcp-admin)

## Channel facts

- Endpoint: `https://go.coremodels.io/mcp`
- Transports: `streamable-http`
- Auth: `required`
- Version: `1.1.0`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically against the live server, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-11-21.

- **Endpoint Security**: 91/100
  - The endpoint's TLS certificate is valid, in date, and uses a strong key.
  - Authorisation is enforced on tool calls, advertised via RFC 9728 protected-resource metadata. Discovery is public, which costs nothing: no tool can be invoked without a token.
  - HTTPS is enforced; there's no plaintext access path.
  - HSTS check failed: the Strict-Transport-Security header is absent.
  - DNSSEC is configured correctly; the domain's records validate against the full chain to the root.
  - The authorisation server offers only Dynamic Client Registration (RFC 7591), which MCP 2026-07-28 deprecated in favour of Client ID Metadata Documents.
- **Transport & Reachability**: 100/100
  - Verified streamable-http transport via a live MCP handshake.
- **Schema Quality & AI Usability**: 56/100
  - AI-judged instruction clarity (good).
  - Context-footprint check failed: tool/resource definitions use about 8143 tokens (~220/item across 37 items; 37 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 0/100
  - Stability check failed: schema churn in the 4 days we've observed: 6 tool removals, 0 breaking changes, 0 auth/transport breaks, 2 additions.
- **Tool Coverage**: 86/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 59% of tool parameters carry a description.
- **Tool Safety**: 75/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - 0 of 1 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation; "run_code" implies "execute" and declares readOnlyHint instead, contradicting what its own name says it does.
  - An AI judge read all 38 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a supported MCP spec version (2025-11-25); the latest is 2026-07-28.

## Install

### How do I install the CoreModels MCP server?

CoreModels is a hosted endpoint at https://go.coremodels.io/mcp, so there is nothing to install locally. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add --transport http io-coremodels-coremodels 'https://go.coremodels.io/mcp'
```

### Cursor

```json
{
  "mcpServers": {
    "io-coremodels-coremodels": {
      "url": "https://go.coremodels.io/mcp"
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "io-coremodels-coremodels": {
      "type": "http",
      "url": "https://go.coremodels.io/mcp"
    }
  }
}
```

### Codex

```toml
[mcp_servers.io-coremodels-coremodels]
url = "https://go.coremodels.io/mcp"
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "io-coremodels-coremodels": {
      "type": "remote",
      "url": "https://go.coremodels.io/mcp",
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add io-coremodels-coremodels --url 'https://go.coremodels.io/mcp' --transport streamable-http
```

### Hermes

```yaml
mcp_servers:
  io-coremodels-coremodels:
    url: "https://go.coremodels.io/mcp"
```

### Netclaw

```json
{
  "McpServers": {
    "io-coremodels-coremodels": {
      "Transport": "http",
      "Url": "https://go.coremodels.io/mcp"
    }
  }
}
```

### Vellum

```bash
assistant mcp add io-coremodels-coremodels -t streamable-http -u 'https://go.coremodels.io/mcp'
```

### Other

```json
{
  "mcpServers": {
    "io-coremodels-coremodels": {
      "type": "http",
      "url": "https://go.coremodels.io/mcp"
    }
  }
}
```

The mcpServers block is a cross-client convention. Remote transports vary, so check your client's docs.

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-11-21 (score 72, 0)

- [functional] This server's schema is too large to store in full, so we cannot compare its tools day to day

### 2026-11-20 (score 72, 0)

- [functional] This server's schema is too large to store in full, so we cannot compare its tools day to day

### 2026-11-19 (score 72, 0)

- [security regression] Stability: 0.03 → fail
- [functional regression] Schema quality: 189 → 220
- [functional] This server's schema is too large to store in full, so we cannot compare its tools day to day

### 2026-11-18 (score 72, 0)

- [functional improvement] Stability: unverified → 0.03
- [functional] This server's schema is too large to store in full, so we cannot compare its tools day to day

### 2026-11-17 (score 72, +36)

- [security improvement] Transport: unverified → pass
- [security improvement] Injection markers: unverified → pass
- [security] First check of Judged manipulation: pass
- [security] Authorization: Authorisation is enforced on tool calls, advertised via RFC 9728 protected-resource metadata. Discovery is public, which costs nothing: no tool can be invoked without a token.
- [functional improvement] MCP protocol: unverified → pass
- [functional improvement] Tool coverage: unverified → 100
- [functional] First check of Schema quality: fail
- [functional] First check of Schema quality: fail
- [functional] First check of Destructive annotations: 0
- [functional] First check of Tool coverage: 57
- [functional] First check of Schema quality: good

### 2026-11-16 (score 36)

First indexed and scored.

## MCP tools (37)

### `export_avro` (~68 tokens)

Export Avro

Export project data as an Avro (.avsc JSON) schema string.

Input parameters:

- `graphProjectId` (string, required)
- `nodeIds` (array): Optional type node ids to export; omit to export all types (optionally scoped to a space).
- `spaceId` (string)

### `analyze_concept_impact` (~145 tokens)

Analyze Concept Impact

Answers 'what is affected if this concept changes?' for a governed type, element or controlled vocabulary. Returns: where it is used inside the model (blast radius), which vendor estates carry it and by what path, which external standards it is bound to, and which serializations it can be exported as. Read-only — nothing is modified. Useful before editing a vocabulary term or a column's meaning.

Input parameters:

- `concept` (string, required): The governed type, element or taxonomy — by node id, or by label (case-insensitive), e.g. 'order status values' or 'status'.
- `graphProjectId` (string, required)
- `spaces` (array)

### `audit_vendor_project` (~178 tokens)

Audit Vendor Project

Audits vendor artifacts (e.g. a dbt manifest) against the governed CoreModels graph — read-only. Reports coverage (what is/isn't governed, canonical-mapping coverage), drift (fields added/removed/retyped, enum changes, contract flips), and conformance (vendor best-practice rules). errorCount > 0 is the CI-gate fail signal; 'markdown' holds the human-readable report.

Input parameters:

- `artifactUrls` (object): Artifact name -> URL to fetch it from (for payloads too large to inline).
- `artifacts` (object): Artifact name -> raw content. dbt: manifest (required), catalog, semantic_manifest.
- `graphProjectId` (string, required)
- `spaces` (array): Optional target space ids.
- `vendor` (string, required): Vendor connector key, e.g. 'dbt'.

### `generate_vendor_artifacts` (~250 tokens)

Generate Vendor Artifacts

Generates vendor artifacts from the governed CoreModels graph. For dbt: a schema.yml with enforced model contracts — column data types, not_null/unique constraints and tests, accepted_values tests from governed taxonomies, relationships tests from governed references. SMEs change meaning once in CoreModels; engineers get a reviewed contract diff.

Input parameters:

- `graphProjectId` (string, required)
- `iriInDescription` (boolean): dbt: also append bound ontology IRIs to the column description so persist_docs carries them into the warehouse column comment. Default true; they always ride in meta regardless.
- `layout` (string): dbt file layout: 'model' (default) one property file per model colocated with its .sql; 'folder' one file per model directory; 'single' one repo-wide file.
- `spaces` (array)
- `targetVersion` (string): Vendor dialect/version. dbt: default emits data_tests: (>=1.8); '1.7' emits legacy tests:.
- `typeNames` (array): Restrict generation to these model/type names (empty = everything eligible).
- `vendor` (string, required): Vendor connector key, e.g. 'dbt'.

### `get_vendor_integration_status` (~92 tokens)

Vendor Integration Status

Lists the registered vendor connectors (dbt, ...) with capabilities and expected artifacts, and — when 'vendor' is given — the project's last-import state for that vendor (when, tool version, artifact versions, fingerprint, counts, governed dataset count).

Input parameters:

- `graphProjectId` (string, required)
- `spaces` (array)
- `vendor` (string): Omit to list all registered vendor connectors.

### `export_json_ld` (~231 tokens)

Export JSON-LD

Export project data in JSON-LD format using a configured export profile.

Use `fetch_json_ld_import_profiles` first to discover the configTypeId. At least one of exportTypes/exportElements/exportTaxonomies/exportDataTypes must be true. Tree-based mode (graphBased=false) requires exactly one nodeId; graph-based mode allows multiple or none.

Input parameters:

- `configTypeId` (string, required): JSON-LD export profile id (from fetch_json_ld_import_profiles)
- `exportDataTypes` (boolean)
- `exportElements` (boolean)
- `exportTaxonomies` (boolean)
- `exportTypes` (boolean)
- `graphBased` (boolean): true = graph-based (no root); false = tree-based (requires exactly one nodeId)
- `graphProjectId` (string, required)
- `includeSpace` (boolean)
- `nodeIds` (array): For tree-based export, exactly one root node id; for graph-based, optional
- `spaceId` (string): Optional space id to filter export data; empty = all spaces
- `useDefaultType` (boolean)

### `fetch_json_ld_import_profiles` (~151 tokens)

Fetch JSON-LD Import Profiles

Fetch JSON-LD import/export profiles available for a project (id and label each).
Pass configTypeId to read that one profile's field mapping instead: every JSON-LD key it knows and what the key imports as - importAs is mixin (target = mixin column id), relation (target = relation group id), newMixin/newRelation (created on import), space, or none (ignored). targetLabel names the mixin column or relation group; isReversed marks a relation read in the opposite direction.

Input parameters:

- `configTypeId` (string): Optional. A profile id from this tool's list; returns that profile's field mapping.
- `graphProjectId` (string, required)

### `export_jsonschema` (~188 tokens)

Export JSON Schema

Export project data as a JSON Schema string.

Input parameters:

- `configTypeId` (string): Export profile id used for the JSON Schema export.
- `graphProjectId` (string, required)
- `inheritanceHandler` (string): How Inheritance relations are emitted: None ignores them, IncludeInProperties flattens inherited elements into the child's properties, IncludeAsRef emits the parent under $defs and references it.
- `inheritenceHandler` (string): Deprecated misspelling of 'inheritanceHandler', kept for existing callers. Use 'inheritanceHandler'; passing both with different values is an error.
- `rootNodeId` (string): Root node id (a Type). Optional only when the project has one Type or one node carries the Root Id mixin; with several Types it is required, and a call without it is refused with the list of Types.
- `spaceId` (string)

### `fetch_json_schema_import_profiles` (~156 tokens)

Fetch JSON Schema Import Profiles

Fetch JSON Schema import/export profiles available for a project (id and label each).
Pass configTypeId to read that one profile's field mapping instead: every JSON Schema keyword it knows and what the keyword imports as - importAs is mixin (target = mixin column id), relation (target = relation group id), newMixin/newRelation (created on import), space, or none (ignored). targetLabel names the mixin column or relation group; isSpecialKey marks a structural JSON Schema keyword (properties, items, $ref, ...).

Input parameters:

- `configTypeId` (string): Optional. A profile id from this tool's list; returns that profile's field mapping.
- `graphProjectId` (string, required)

### `validate_json` (~347 tokens)

Validate JSON

Validate one or more JSON documents against a project's stored JSON Schema.

The schema is regenerated from the project using the supplied configTypeId (the export profile id) and rootNodeId, then reused for every document - so validating a batch costs barely more than validating one. Pass up to 100 documents per call and chunk anything larger.

Response:
\- isValid: true only when EVERY document passed.
\- total, validCount: counts over the whole batch.
\- invalid: one entry per failing document, as {index, errors}. Only failures are listed - any index not present passed. index is the 0-based position in the jsonStrings array you sent.
\- A document that is not parseable JSON fails in its own slot; it does not abort the rest of the batch.

Set explain=true to add a plain-English "- Explanation:" line to every error, describing what is wrong, where, and how to fix it. The raw structured lines are unchanged either way - use it when a person will read the result, leave it off for machine parsing.

Input parameters:

- `configTypeId` (string): Profile id used to map mixins/relations to schema keywords
- `explain` (boolean): Add a plain-English explanation to each error alongside the raw structured fields. Use when a person will read the output.
- `graphProjectId` (string, required)
- `jsonStrings` (array, required): The JSON documents to validate, each serialized as a string. Results reference documents by their 0-based position in this array.
- `rootNodeId` (string, required): Node id used as the root when generating the JSON Schema
- `spaceId` (string): Optional space id to scope the schema

### `export_linkml` (~67 tokens)

Export LinkML

Export project data as a LinkML (YAML) schema string.

Input parameters:

- `graphProjectId` (string, required)
- `nodeIds` (array): Optional type node ids to export; omit to export all types (optionally scoped to a space).
- `spaceId` (string)

### `check_admissible` (~267 tokens)

Check Admissible

Check whether a relation would be admissible before writing it. Reads the SAME relation-group
rule table create_relation and bulk_create validate against, so the answer matches what a write
would do.

Pass relationGroupId plus fromNodeId and toNodeId — both must already exist in the project.
Returns admissible, the rule (relation group label) it was checked against, a reason when
refused (naming the allowed direction), and alternative relation groups that do admit this
node-type pair.

Mixin form: instead of the relation arguments, pass mixinId, targetNodeId and mixinColumns (the same
items bulk_create takes under mixins; id = column ID, label = value). The answer comes from the same
dry-run checks bulk_create runs for mixin values, with nothing written: admissible, and the reason when
refused. Do not pass both forms in one call.

Input parameters:

- `fromNodeId` (string)
- `graphProjectId` (string, required)
- `mixinColumns` (array): id = column ID, label = value.
- `mixinId` (string)
- `relationGroupId` (string)
- `targetNodeId` (string)
- `toNodeId` (string)

### `list_comments` (~311 tokens)

List Comments

List the discussion (comments and replies) of a CoreModels project — the same threads users see in the app's grid collaboration panel. Viewer access is required.

Comments are anchored by tag sets: a comment left on a grid cell carries that cell's tag ids (node/element/relation ids, sometimes a section label) plus the project id. Omit anchorIds to list the whole project's discussion; pass anchorIds to return only comments anchored to exactly that tag set (the project id is implied — never include it).

Returns { comments, total, returned, truncated }. comments are top-level threads, newest first, each with its replies nested oldest first; every entry has { id, createdBy, createdByName, message, anchorIds, mentionUserIds, parentId, isResolved, createdDate }. Reply to a thread or leave a new one with add_comment (admin endpoint).

Input parameters:

- `anchorIds` (array): Optional cell anchor: return only comments whose anchor tag set equals exactly this set. Do not include the project id.
- `graphProjectId` (string, required): Id of the project whose discussion to read. Viewer access is required.
- `includeReplies` (boolean): Nest each thread's replies in the response.
- `includeResolved` (boolean): Also return resolved (closed) threads.
- `maxResults` (integer): Cap on returned top-level threads (newest first); truncated=true when more matched.
- `textContains` (string): Optional text filter, matched in comments and replies.

### `get_job_status` (~177 tokens)

Get Job Status

Get the status of a background job in a project, e.g. the copy started by copy_project or a background import_owl. Returns { jobDetailsId, projectId, type, status, createdAt, lastUpdatedAt } where status is "inProgress", "success" or "failed". Result-bearing jobs (background import_owl) additionally return a 'result' field on success — for OWL imports it carries the type/element/taxonomy counts and the aggregated lossiness. Poll until status is no longer "inProgress"; leave a few seconds between polls.

Input parameters:

- `graphProjectId` (string, required): Id of the project the job belongs to. For copy_project this is the returned (destination) projectId.
- `jobDetailsId` (string, required): The jobDetailsId returned by the tool that started the job.

### `get_mixins_and_relation_groups` (~104 tokens)

Get Mixins & Relation Groups

Get the project schema: all mixin definitions and all relation-group definitions.

Use this once at the start of a session to discover the IDs needed by other tools (mixinId, columnId, relationGroupId). Returns compact positional arrays - see the "format" field for the layout.

Read get_modeling_guide before your first write - it explains what each relation group means and which direction it runs.

Input parameters:

- `graphProjectId` (string, required)

### `get_modeling_guide` (~78 tokens)

Get Modeling Guide

Returns the CoreModels modeling guide (markdown): how to import vs build, what each relation group means and which direction it runs (Domain Includes, Range Includes, Controlled List, Has List Item, Inheritance, Derived From), bulk_create conventions, and how to verify your work. Read it once before your first write in a session.

### `get_project_summary` (~107 tokens)

Get Project Summary

Labels and IDs of types, elements, and taxonomies in the project. Each category is paginated independently.

Pagination:
\- First call: omit page (defaults to 0).
\- Each category reports page, pageSize and hasMore. Re-request with page+1 for any category whose hasMore is true.

Input parameters:

- `graphProjectId` (string, required)
- `page` (integer): 0-based page index, applied independently to each category.
- `pageSize` (integer)

### `list_projects` (~284 tokens)

List Projects

List the user's CoreModels projects as [id,name,accessLevel] (see the response "format" field). Use a returned id as graphProjectId for other tools. Pass searchTerm to filter by name (case-insensitive substring). Set includePublicProjects=true to also include public projects. Set includeAISummary=true to also return each project's saved AI-generated summary and the time it was generated (4th and 5th elements). The response's inLibrary maps each returned id to whether the project is in the Schema Library (schema_library_projects_section); a public project need not be curated into the library, and external library catalog entries are not graph projects listed here. Paged: page is 1-based; increment page up to the returned totalPages to get all results.

Input parameters:

- `includeAISummary` (boolean): When true, each project row includes its saved AI-generated summary (markdown) and the time it was generated as 4th and 5th elements, or null if none has been generated.
- `includePublicProjects` (boolean): When true, public projects the user is not a member of are also included.
- `page` (integer): 1-based page number. Increment to page through results up to totalPages.
- `pageSize` (integer)
- `searchTerm` (string): Optional case-insensitive substring to filter projects by name.

### `core_models_get_mixins_info` (~80 tokens)

Core Models Get Mixins Info

Get information about all mixins in a CoreModels project.

Mixins define structured metadata used for mapping models, system integrations, and other extensions.
Mixin values are attached to nodes; use search_nodes with includeMixins to see them, and bulk_create or create_mixin_value to set them.

Input parameters:

- `graphProjectId` (string, required)

### `query_model_graphql` (~398 tokens)

Query Model Graphql

Run a read-only GraphQL query against a CoreModels project's model — the general query escape hatch for precise questions (projections, traversals, counts) that the fixed read tools don't answer. Viewer access is required, same as the app.

The schema is GENERATED PER PROJECT: field names are camelCased from this project's labels and cannot be guessed. Always call with schemaOnly=true first and write queries against the returned SDL (standard introspection queries also work, but the SDL is far cheaper). Root fields: types, elements, taxonomies, views, exemplars, spaces, mixins, relationGroups. Node roots take filter arguments id, ids, names, spaceId — note `names` is a search expression, not an exact match.

Queries only: the schema has no Mutation or Subscription type. GraphQL variables and operationName are not supported — inline all literals into the query document.

There is NO pagination and every matched node is returned with all its relations, so select narrowly and always filter. Results whose serialized data exceeds maxChars come back as { truncated: true, totalChars, maxChars, dataPreview } instead of { data } — tighten the query rather than raising maxChars. On a GraphQL error (usually a wrong field name) you get an error result; re-check the schema with schemaOnly=true.

Input parameters:

- `graphProjectId` (string, required): Id of the project to query. Viewer access is required.
- `maxChars` (integer): Cap on the serialized result size. The server has no pagination; over the cap you get truncated:true with a string preview instead of data.
- `query` (string): GraphQL query document (query operations only; no variables — inline all literals). Required unless schemaOnly is true.
- `schemaOnly` (boolean): Return the project's generated GraphQL schema as SDL without executing anything. Call this first: field names are camelCased from this project's labels.

### `core_models_get_relation_groups_info` (~44 tokens)

Core Models Get Relation Groups Info

Get information about relation groups for a CoreModels project.

Use the relation groups to create a relation between two nodes.

Input parameters:

- `graphProjectId` (string, required)

### `search_nodes` (~423 tokens)

Search Nodes

Search nodes in a CoreModels project. Returns compact positional arrays; the response "format" field describes the layout.

Filters (provide at least one of the first three; they combine with AND):
\- nodeIds: exact id lookup
\- nodeType: one of Element, Type, Taxonomy, Exemplar, Component, Space, Tag, Mixin
\- expression: partial substring match on the node label (plain text, no wildcards)
\- spaceIds: restrict results to specific spaces. Narrowing only - cannot be the sole filter; always combine it with nodeIds, nodeType, or expression.

Optional flags: includeRelations, includeMixins, includeTimestamps, sortAttr, sortDesc, pageSize.

Pagination:
\- Page-number based: pass page (1-based) and pageSize (max 250).
\- The response includes the total match count, so more pages exist while page * pageSize < total.

Input parameters:

- `expression` (string): Partial substring match against the node label. Plain text only - no wildcards, no regex, no '*', '%', '_' or '?' characters; the literal characters are matched as-is.
- `graphProjectId` (string, required)
- `includeMixins` (boolean)
- `includeRelations` (boolean)
- `includeTimestamps` (boolean): Adds creationTime and updateTime (ISO 8601 UTC) to each node. Combine with sortAttr CreationTime/UpdateTime to see when nodes changed.
- `limitPerRelationGroup` (integer): Only with includeRelations. Caps relations returned PER relation group for each node (e.g. 100). Omit for no cap.
- `nodeIds` (array)
- `nodeType` (string)
- `page` (integer): 1-based page number. The response includes 'total' (the total match count) so you can compute how many pages there are.
- `pageSize` (integer)
- `sortAttr` (string)
- `sortDesc` (boolean)
- `spaceIds` (array)

### `whoami` (~253 tokens)

Whoami

Who is calling: returns the identity and subscription plan of the MCP token in use. No project id needed.

Returns { userId, name, email, organizationId, endpoint, plan }. endpoint is { path, mode (read-only or write), writeEndpoint } for the endpoint this call arrived on. plan holds the raw tier and expiry from the token plus the derived limits and permissions the other tools enforce: canWrite (writes allowed at all), sharingAllowed (whether invite_user will work — the app's Share-button rule), isEnterprise (large-project copies), maxProjects (null when no limit is enforced on MCP, with maxProjectsEnforced=false; planMaxProjects is the web-app plan value) and maxNodesPerProject (same: absent when not enforced, with maxNodesEnforced=false, maxNodesNote and planMaxNodesPerProject). limitsEnforced says whether those two caps are applied on MCP calls now, and limitsEnforcedFrom when they start (2026-12-31; the sandbox project is exempt); while limitsEnforced is false, maxProjects and maxNodesPerProject stay null as above. Call it at the start of a session to know your limits before creating or importing.

### `suggest_ontology_terms` (~326 tokens)

Suggest Ontology Terms

Suggest ontology-term candidates for a schema node from the public EMBL-EBI Ontology Lookup Service
(OLS). ADVISORY ONLY: this is a read-only search — it calls the external EBI OLS API and returns
ranked candidates, and it NEVER writes anything to the graph. In particular it never sets the
authoritative mapsTo mixin; binding a chosen term is a separate, explicit human-directed act via
bind_ontology_term (you name the exact node and the exact term IRI).

Provide either nodeId (the node's label is used as the query) or a raw label (optionally with a
description). Optionally restrict to a comma-separated list of ontologies (e.g. "efo,mondo,hp") and
cap the number of candidates. Each candidate carries { iri, curie, label, description, ontology, type }
in the relevance order OLS returns; the iri is what you would pass to bind_ontology_term.

Input parameters:

- `description` (string): Optional description text to refine the search ranking.
- `graphProjectId` (string, required)
- `label` (string): A raw term label to search for. Provide this OR nodeId.
- `limit` (integer): Maximum candidates to return (1-50).
- `nodeId` (string): The schema node to ground. Its label is used as the query. Provide this OR label.
- `ontologies` (string): Optional comma-separated ontology ids to restrict to, e.g. 'efo,mondo,hp'.

### `export_shex` (~245 tokens)

Export ShEx

Export project data as a ShEx (Shape Expressions) schema string.

Input parameters:

- `closed` (boolean): When true, every Type shape is CLOSED (no properties beyond the ones listed). Taxonomy shapes stay open. When false, a shape with a type constraint gets 'EXTRA rdf:type' so the node may have other ty…
- `graphProjectId` (string, required)
- `includeCardinality` (boolean): Whether to emit ShEx cardinality markers (?, *, +). Defaults to true.
- `linkMode` (string): How an element whose range is another Type is written: 'shape' writes @<OtherShape> (the linked node must also match that shape), 'iri' writes IRI (the linked node only has to be an IRI).
- `nodeIds` (array): Optional type node ids to export; omit to export all types (optionally scoped to a space).
- `spaceId` (string)
- `typeConstraint` (boolean): When true, a shape whose Type is bound to a class (bind_ontology_term / mapsTo) gets 'rdf:type [ <classIRI> ]'. Types with no binding are unchanged.

### `export_sql` (~186 tokens)

Export SQL

Export the project's types as runnable SQL DDL (CREATE TABLE statements) for PostgreSQL / SQLite.

Identifiers are double-quoted; column types come from the Json Schema Mixin (number -> NUMERIC, integer -> INTEGER, boolean -> BOOLEAN, string date/date-time -> DATE/TIMESTAMP, else VARCHAR); required -> NOT NULL; every table gets a PRIMARY KEY (the element marked key, else a generated "id" VARCHAR(36)); a Controlled List -> CHECK (col IN (...)); a Range Includes to a Type -> a foreign key (on the child table for an array, e.g. "customer_id" on "Order"); referenced tables come first.

Input parameters:

- `graphProjectId` (string, required)
- `nodeIds` (array): Optional type node ids to export; omit to export all types (optionally scoped to a space).
- `spaceId` (string)

### `export_owl` (~123 tokens)

Export OWL (Turtle)

Export project schema as an OWL ontology (Turtle): owl:Class per Type, Datatype/ObjectProperty per
Element with domain/range, rdfs:subClassOf inheritance, cardinality restrictions preserving
required/array-ness, SKOS concept schemes per Taxonomy, and owl:equivalentClass/Property from
cross-standard mappings. Anything OWL cannot express is reported as '# lossiness:' comments at the
top of the document instead of being silently dropped.

Input parameters:

- `graphProjectId` (string, required)
- `spaceId` (string)

### `export_rml` (~369 tokens)

Export RML mapping (Turtle)

Export the project as an RML mapping document (RML-Core + RML-IO, namespace http://w3id.org/rml/):
one rml:TriplesMap per Type, an rml:logicalSource per source file or table, a subject template from
the identifier element, one rml:predicateObjectMap per Element, and joins (rml:parentTriplesMap +
rml:joinCondition) for type references. Run the result in an RML engine (RMLMapper, morph-KGC,
SDM-RDFizer) to build a knowledge graph from your data.

Give sourceKind to bind it to real data (csv | json | xml | sql | sql-query); omit it for a
skeleton with placeholder sources. The document's first line states its fidelity: RUNNABLE,
TEMPLATE or SKELETON.

Class and predicate IRIs come from each Type's and Element's mapsTo binding. Where a term is not
bound, the IRI is minted from its name and reported as '# lossiness:' — a mapping built on minted
IRIs produces a graph that joins to nothing, so bind the terms (bind_ontology_term) before
publishing the output.

Input parameters:

- `baseIri` (string): Namespace subject IRIs are built under, e.g. https://data.example.gov/id
- `graphProjectId` (string, required)
- `sourceForm` (string): shorthand (default) is what RML engines accept today; rml-io is the newer structured source form
- `sourceKind` (string)
- `sourceTemplate` (string): Path or table per type, with {type} substituted, e.g. data/{type}.csv
- `spaceId` (string)

### `generate_synapse_manifests` (~226 tokens)

Generate Synapse Manifests

Generate Synapse curation-manifest CSV templates from a schema — the tabular companion to the
transform_schema synapse export. Two artifacts per type: {name}.manifest.csv (the blank entry grid,
header row only, loadable as a Curator RecordSet CSV) and {name}.dictionary.csv (the column
dictionary: Column, Type, Description, Required, Value Set — followed by a value-set section listing
taxonomy terms for dropdown columns). Source formats: jsonschema | shex | avro | jsonld | sql | osi |
osi-json | owl | linkml | odm. Stateless — the project scopes auth only; nothing is written.

Input parameters:

- `graphProjectId` (string, required)
- `sourceFormat` (string, required): jsonschema | shex | avro | jsonld | sql | osi | osi-json | owl | linkml | odm
- `sourceSchema` (string, required): The source schema text.
- `typeId` (string): Generate manifests for this type only (matched by id or label). Omit for every type.

### `transform_schema` (~686 tokens)

Transform Schema

Transform a schema from one format to another through the ARAMAI mapping engine.
Formats: jsonschema | shex | avro | jsonld | sql | osi | osi-json | owl | linkml | protobuf | odcs | graphql | odm | synapse
(odm decodes only; osi-json is the OSI coder in its JSON serialization — the form dbt ingests;
synapse encodes only — the Sage Bionetworks Synapse draft-07 subset with a registered-schema $id;
its output is plain JSON Schema, decodable via jsonschema; protobuf is a single proto3 .proto file,
alias proto; odcs is a Bitol Open Data Contract Standard v3 YAML data contract; graphql is a GraphQL
SDL type-system document, alias sdl — the universal fallback for importing any CMS content model,
since every headless CMS can emit an SDL schema; decode imports the content model and skips the
generated query surface — root operation types, Connection/Edge/PageInfo pagination, filter inputs —
with an honest lossiness summary).
Mapping: 'inferred' (label/type matching against a target hint), 'explicit' (a SIA mapping-guide
JSON: autoMatchByMapsTo, fieldMappings [{sourceElementIds, targetElementIds, transformName}],
taxonomyDirectives, drops), or 'ai' (Claude proposes a plan; the same validation gate accepts or
rejects it — requires a server-configured Anthropic key and Editor/Admin project membership, and
sends the schemas to the Anthropic API server-side). The response carries the produced schema,
the executed plan (replayable), and an honest lossiness ledger of everything the transformation
could not preserve.

Input parameters:

- `caseInsensitive` (boolean): Inferred mapping: ignore label case (default true).
- `graphProjectId` (string, required)
- `guide` (string): For explicit mapping: the SIA mapping guide as JSON text. For ai mapping: optional free-text guidance for the model.
- `mappingKind` (string): inferred (default) | explicit | ai
- `sourceFormat` (string, required): jsonschema | shex | avro | jsonld | sql | osi | osi-json | owl | linkml | protobuf | odcs | graphql | odm
- `sourceSchema` (string, required): The source schema text.
- `synapseName` (string): For synapse output: the schema name segment of the $id (default: derived from the root type).
- `synapseOrg` (string): For synapse output: the Synapse Organization name for the $id (default 'example'; org.sagebionetworks is reserved).
- `synapseVersion` (string): For synapse output: the semver segment of the $id (default 0.0.1; Synapse rejects re-registering an existing version).
- `targetFormat` (string, required): jsonschema | shex | avro | jsonld | sql | osi | osi-json | owl | linkml | protobuf | odcs | graphql | synapse
- `targetHintFormat` (string): Format of the optional target hint schema.
- `targetHintSchema` (string): The schema to map toward. Required for inferred mapping.
- `vendor` (string): For sql output: postgres (default) | mysql | sqlserver.

### `schema_library_query_project` (~63 tokens)

Schema Library Query Project

Query and retrieve details of a SchemaLibrary project including metadata, classifications, likes, cover images, and user permissions. Authenticated callers receive additional user-specific data (e.g. ILiked, Role).

Input parameters:

- `projectId` (string, required): The SchemaLibrary project ID

### `schema_library_get_metadata_options` (~64 tokens)

Schema Library Get Metadata Options

Get available metadata options for SchemaLibrary projects (access methods, types, licenses, formats, domains, audiences, model sizes). Use these IDs when creating or updating projects. Optionally filter by category.

Input parameters:

- `category` (string): Filter by metadata category (optional)

### `schema_library_get_classifications` (~56 tokens)

Schema Library Get Classifications

Get available classification options for SchemaLibrary projects, used to categorize and tag projects. Optionally filter by classification name (partial match supported).

Input parameters:

- `nameFilter` (string): Filter by classification name (optional - partial match supported)

### `schema_library_projects_section` (~287 tokens)

Schema Library Projects Section

Search and retrieve a filtered, paginated list of SchemaLibrary projects with advanced filtering options including metadata, classifications, and project attributes. The library is the curated catalog, not the graph-project list: each row's inProjectList says whether list_projects (includePublicProjects=true) also returns it — external catalog entries (isExternal) and suspended or stopped projects are library-only.

Input parameters:

- `accessMethodId` (string)
- `accessTypeId` (string)
- `audienceIds` (string): Comma-separated audience IDs
- `classificationIds` (string): Comma-separated classification IDs
- `domainIds` (string): Comma-separated domain IDs
- `formatIds` (string): Comma-separated format IDs
- `graphProjectType` (string): Graph project type filter (e.g. 'Public', 'Private')
- `isExternal` (boolean)
- `isFeatured` (boolean)
- `licenseId` (string)
- `modelSizeId` (string)
- `onlyFavoriteProjects` (boolean): Authenticated only
- `onlyMyProjects` (boolean): Authenticated only
- `page` (integer)
- `pageSize` (integer)
- `publishState` (string)
- `sortBy` (string)
- `title` (string): Search by project title (partial match)
- `typeId` (string)

### `schema_library_get_use_case_types` (~34 tokens)

Schema Library Get Use Case Types

Get available use case types for SchemaLibrary projects. Use these types when creating new use cases for a project.

### `contact_sales` (~319 tokens)

Contact Sales

Ask the CoreModels team for a scoping call: the escalation path when this MCP server cannot do the
job alone (a plan limit, a missing capability, or a question only a person can answer). Call it when
you are stuck for that reason, not as a first resort.

Creates a sales lead; a person from the CoreModels team follows up by email. No project id and no
sign-in needed: this is the one tool on /mcp that works without a JWT. Rate-limited per caller and
in total. The person named in contact_email must have agreed to be contacted (consent: true).

Input parameters:

- `calling_agent_name` (string): Name of the AI agent or application making this call (self-reported, e.g. 'Claude Desktop', 'Cursor', a custom script name).
- `company` (string, required): The contact's company or organization.
- `consent` (boolean, required): Must be true: confirms the person named in contact_name/contact_email has agreed to be contacted by email.
- `contact_email` (string, required): Email of the person who should be contacted. They are told why: a person replies within one business day.
- `contact_name` (string, required): Name of the person who should be contacted.
- `need` (string, required): What help is needed, in your own words — the plan limit, missing capability, or question that brought you here.
- `recent_tool_calls` (array): Names of the MCP tools you called before this one, in order, for context on what led here.

### `run_code` (~331 tokens)

Run Code

Execute a JavaScript program that orchestrates this server's tools, and return only its result.

Prefer this over many individual tool calls when a task needs several steps, looping, filtering, or
combining data: intermediate results stay in the sandbox, so only what you return reaches the model.

Inside the script:
\- listTools() -> [{name, summary}]            discover available tools
\- getToolDoc(name) -> {name, description, parameters, required}   inspect one tool's inputs
\- tools.<name>(args) -> parsed result          call a tool (graphProjectId is injected automatically; do NOT pass it)
\- console.log(...)                             captured and returned alongside the result
\- return <value>                               JSON-serialized and returned

Environment: sandboxed JavaScript, no network or filesystem. Limits: 30 seconds, 128 MB of memory,
200,000 statements, 50 tool calls per script (20 of them writes on the admin endpoint).
This endpoint serves read-only tools, so only read tools are ever callable from code here. Code-mode
writes (an allowlist of creation tools) are served by the admin endpoint (/mcp-admin).
Use listTools() to see what is callable right now.

Input parameters:

- `code` (string, required): JavaScript to execute. Use tools.<name>(args), listTools(), getToolDoc(name); end with `return <value>`.
- `dryRun` (boolean): When true, write-tool calls are collected and reported as plannedWrites instead of being executed. Read tools run normally.
- `graphProjectId` (string, required)

## Diagnostics

Captured diagnostic sections: TLS, DNSSEC, Authorisation, Transports. The full working is on the page: https://verifymcp.io/servers/io-coremodels-coremodels/go#diagnostics

## Score history

- 2026-11-21: 72
- 2026-11-20: 72
- 2026-11-19: 72
- 2026-11-18: 72
- 2026-11-17: 72
- 2026-11-16: 36

## Common questions

### What is the CoreModels MCP server?

CoreModels is an MCP server listed in the public MCP registry as io.coremodels/coremodels. Model, map and validate data schemas. Export JSON Schema, JSON-LD, LinkML, OWL, ShEx, SQL, Avro. This page covers its hosted endpoint (https://go.coremodels.io/mcp).

### Is the CoreModels MCP server safe to use?

CoreModels scores 72 out of 100 on VerifyMCP. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the CoreModels MCP server expose?

CoreModels exposes 37 tools: export_avro, analyze_concept_impact, audit_vendor_project, generate_vendor_artifacts, get_vendor_integration_status, and 32 more. Their descriptions and schemas cost roughly 7,714 tokens of context every time the server is loaded.

### Does the CoreModels MCP server require authentication?

Yes. CoreModels asked us for credentials when we connected, so you will need to authorise it in your MCP client before it can do anything.

### Is the CoreModels MCP server still maintained?

CoreModels is still listed as active in the MCP registry. We last reached this channel on 21 November 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

### What did VerifyMCP test on the CoreModels MCP server?

VerifyMCP ran 13 checks on CoreModels, covering TLS certificate, Authorization, HTTPS, HSTS header and 9 more. Supply Chain Security and Provenance & Transparency do not apply to a hosted endpoint, so they were not tested. The score does not measure everything, and the scoring documentation lists what it leaves out.

## Links

- Remote endpoint: https://go.coremodels.io/mcp
- Website: https://www.coremodels.io/mcp-server
- Changelog RSS feed: https://verifymcp.io/servers/io-coremodels-coremodels/go.xml
- Changelog JSON feed: https://verifymcp.io/servers/io-coremodels-coremodels/go/changelog.json
- Trust report (JSON): https://verifymcp.io/servers/io-coremodels-coremodels/go.json
- HTML version of this page: https://verifymcp.io/servers/io-coremodels-coremodels/go
