# three.ws for Grok (remote · three.ws)

Free text/image to 3D for Grok Bot, plus your three.ws agents once signed in. Never spends.

- Trust score: 79/100 (medium)
- Change this week: +3
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-10-23

## Components

- remote · `three.ws`: 79/100 (this document), [markdown](https://verifymcp.io/servers/nirholas-threews-grok/api-mcp-grok.md), [page](https://verifymcp.io/servers/nirholas-threews-grok/api-mcp-grok)

## Channel facts

- Endpoint: `https://three.ws/api/mcp-grok`
- Transports: `streamable-http`
- Auth: `none`
- Version: `1.0.0`

## 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 2026-10-23.

- **Endpoint Security**: 83/100
  - The endpoint's TLS certificate is valid, in date, and uses a strong key.
  - No authorisation is required to call this server. Every tool declares its destructiveHint and none is destructive, so open access doesn't expose one.
  - HTTPS is enforced; there's no plaintext access path.
  - The HSTS (Strict-Transport-Security) header is present.
  - DNSSEC is configured correctly; the domain's records validate against the full chain to the root.
- **Transport & Reachability**: 100/100
  - Verified streamable-http transport via a live MCP handshake.
- **Schema Quality & AI Usability**: 79/100
  - 100% of prompts and resources have a non-trivial description (not blank, and not just the item's name).
  - AI-judged instruction clarity (excellent).
  - Context-footprint check failed: tool/resource definitions use about 3691 tokens (~246/item across 15 items; 15 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 37/100
  - Stability observed for 11 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 99/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 98% of tool parameters carry a description.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 15 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 16 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 60/100
  - Spec-recency check failed: implements MCP spec 2025-06-18; the latest is 2026-07-28.

## Install

### How do I install the three.ws for Grok MCP server?

three.ws for Grok is a hosted endpoint at https://three.ws/api/mcp-grok, 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 nirholas-threews-grok 'https://three.ws/api/mcp-grok'
```

### Cursor

```json
{
  "mcpServers": {
    "nirholas-threews-grok": {
      "url": "https://three.ws/api/mcp-grok"
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "nirholas-threews-grok": {
      "type": "http",
      "url": "https://three.ws/api/mcp-grok"
    }
  }
}
```

### Codex

```toml
[mcp_servers.nirholas-threews-grok]
url = "https://three.ws/api/mcp-grok"
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "nirholas-threews-grok": {
      "type": "remote",
      "url": "https://three.ws/api/mcp-grok",
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add nirholas-threews-grok --url 'https://three.ws/api/mcp-grok' --transport streamable-http
```

### Hermes

```yaml
mcp_servers:
  nirholas-threews-grok:
    url: "https://three.ws/api/mcp-grok"
```

### Netclaw

```json
{
  "McpServers": {
    "nirholas-threews-grok": {
      "Transport": "http",
      "Url": "https://three.ws/api/mcp-grok"
    }
  }
}
```

### Vellum

```bash
assistant mcp add nirholas-threews-grok -t streamable-http -u 'https://three.ws/api/mcp-grok'
```

### Other

```json
{
  "mcpServers": {
    "nirholas-threews-grok": {
      "type": "http",
      "url": "https://three.ws/api/mcp-grok"
    }
  }
}
```

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.

### 2026-10-23 (score 79, +1)

No change was recorded against any check on this day. Stability & Change Management went from 33 to 37. That category is still filling its 30-day observation window: 10 days of observed history at the previous scan, 11 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-10-20 (score 78, +1)

No change was recorded against any check on this day. Stability & Change Management went from 23 to 27. That category is still filling its 30-day observation window: 7 days of observed history at the previous scan, 8 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-10-18 (score 77, +1)

No change was recorded against any check on this day. Stability & Change Management went from 17 to 20. That category is still filling its 30-day observation window: 5 days of observed history at the previous scan, 6 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-10-16 (score 76, +1)

No change was recorded against any check on this day. Stability & Change Management went from 10 to 13. That category is still filling its 30-day observation window: 3 days of observed history at the previous scan, 4 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-10-14 (score 75, +1)

No change was recorded against any check on this day. Stability & Change Management went from 3 to 7. That category is still filling its 30-day observation window: 1 days of observed history at the previous scan, 2 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-10-13 (score 74, 0)

- [functional improvement] Stability: unverified → 0.03

### 2026-10-12 (score 74)

First indexed and scored.

## MCP tools (15)

### `forge_free` (~257 tokens)

Generate a 3D model from text

Turn a text prompt into a textured, downloadable 3D model (GLB), free. Describe a single object, character, or creature; the studio generates an interactive model you can rotate, view, and download. Optional quality tier (draft, standard, high); high is slower and may fall back to standard under load. Renders inline in an interactive 3D viewer. Use this for a single object or creature from text; for a posable character call forge_avatar, and to change a model you already made call refine_model.

Input parameters:

- `idempotency_key` (string): Optional retry key. Send the same key again (same caller, same arguments, within 24 hours) and you get the original job back instead of a second generation, so a retry after a timeout cannot burn quo…
- `prompt` (string, required): Description of the single object or character to model, e.g. "a friendly round robot mascot, glossy white plastic".
- `tier` (string): Detail level: draft (fastest), standard (default), or high (best, slower; may fall back to standard under load).

### `text_to_avatar` (~184 tokens)

Generate a 3D avatar

Generate a textured 3D avatar (GLB) from a text description or a reference image URL. Best for characters and figures. Renders inline in an interactive 3D viewer. Use this for a character that only needs to be viewed; when it must be posed or animated call forge_avatar instead.

Input parameters:

- `idempotency_key` (string): Optional retry key. Send the same key again (same caller, same arguments, within 24 hours) and you get the original job back instead of a second generation, so a retry after a timeout cannot burn quo…
- `image_url` (string): Optional http(s) URL to a reference image to reconstruct in 3D.
- `prompt` (string): Description of the avatar to generate.

### `mesh_forge` (~201 tokens)

Generate a 3D mesh (art-directed)

Generate a textured 3D mesh (GLB) from a text prompt or a reference image URL. In text mode an AI art-director first refines your prompt into an optimized single-subject spec for higher mesh quality. Renders inline in an interactive 3D viewer. Use this for a single object or prop when mesh quality matters most; for characters call text_to_avatar or forge_avatar.

Input parameters:

- `idempotency_key` (string): Optional retry key. Send the same key again (same caller, same arguments, within 24 hours) and you get the original job back instead of a second generation, so a retry after a timeout cannot burn quo…
- `image_url` (string): Optional http(s) URL to a reference image to reconstruct directly.
- `prompt` (string): Description of the single object to model.

### `rig_mesh` (~199 tokens)

Rig a 3D model for animation

Auto-rig a static 3D model (GLB) into an animation-ready model: adds a humanoid skeleton and skin weights so it can be posed and animated. Provide the GLB URL of a model (e.g. one generated by the other tools). Renders the rigged result inline in an interactive 3D viewer. Use this when you already have a static model; to generate and rig a character in one step call forge_avatar instead.

Input parameters:

- `glb_url` (string, required): http(s) URL to the static GLB mesh to rig.
- `idempotency_key` (string): Optional retry key. Send the same key again (same caller, same arguments, within 24 hours) and you get the original job back instead of a second generation, so a retry after a timeout cannot burn quo…

### `forge_avatar` (~242 tokens)

Generate a rigged, animation-ready avatar

Generate a rigged, animation-ready 3D avatar (GLB) from a single text prompt or a reference image, in one step. Generates the mesh, then auto-rigs it with a humanoid skeleton so it is ready to pose and animate. Best for characters; objects are steered to the mesh generator. Renders inline in an interactive 3D viewer. Use this when the user wants a character they can pose or animate.

Input parameters:

- `allow_non_humanoid` (boolean): Set true to rig a non-humanoid subject anyway (rigging assumes a humanoid figure).
- `idempotency_key` (string): Optional retry key. Send the same key again (same caller, same arguments, within 24 hours) and you get the original job back instead of a second generation, so a retry after a timeout cannot burn quo…
- `image_url` (string): Optional http(s) URL to a reference image to reconstruct in 3D.
- `prompt` (string): Description of the character/avatar to generate.

### `refine_model` (~408 tokens)

Refine a 3D model by describing a change

Iterate on a model you already generated: just describe the change in words ("make it metallic", "bigger helmet", "add wings"). The studio re-generates a new version anchored to the previous one, carrying its form and materials forward. Pass the previous model's glb_url and, when you have it, the prompt that made it (parent_prompt) so the change builds on it. Each refinement is recorded as a new version in a lineage you can revert to or branch from, and the returned lineage drives a version strip in the viewer. Renders the new version inline in the interactive 3D viewer. Use this instead of a fresh generation when the user wants to change a model they already have.

Input parameters:

- `glb_url` (string, required): http(s) URL of the model to refine (e.g. the glbUrl a previous generation returned).
- `idempotency_key` (string): Optional retry key. Send the same key again (same caller, same arguments, within 24 hours) and you get the original job back instead of a second generation, so a retry after a timeout cannot burn quo…
- `instruction` (string, required): The change to make, in plain language: "make it metallic", "bigger helmet", "add a cape".
- `parent_index` (integer): Optional: branch off an earlier version in parent_lineage (its index) instead of the latest.
- `parent_lineage` (array): Optional: the lineage array from a previous refine_model result, to extend the same version history.
- `parent_prompt` (string): Optional: the prompt that produced the model being refined, so the change builds on it instead of starting over.
- `reference_image_url` (string): Optional http(s) image of the current model to anchor the re-generation (image→3D). Omit for text-guided refinement.

### `check_job` (~158 tokens)

Check a pending 3D generation

Check on a 3D generation that returned status "pending" and collect the finished model. Pass the job_id from the pending result. While it is still rendering you get updated timing; call again after the suggested wait. Checking can save the completed model and restart failed work on another provider. When it is done the model renders inline in the interactive 3D viewer. Use this when the user is waiting to see the finished model; for quiet polling on a schedule call get_job instead.

Input parameters:

- `job_id` (string, required): The job_id (or jobId) a pending generation returned.
- `refine` (object): Optional: the refine object a pending refine_model result carried. Pass it back unchanged so the finished model joins that version history.

### `get_job` (~182 tokens)

Get the status of a 3D job

Machine-readable status of a generation that returned "pending". Returns status (queued, running, done, failed, not_found), progress (0 to 1, null when unknown), eta_seconds, and next_check_seconds while it runs; the asset links (viewer_url, glb_url, poster_png_url, embed_html) when done; and a plain error with a remedy when it failed. Safe to call on a schedule: it never starts work and is not counted against the generation quota. Use this to poll a pending job from code or a loop; call check_job with the same job_id when you want the finished model rendered in the inline viewer.

Input parameters:

- `job_id` (string, required): The job_id (or jobId) a pending generation returned.
- `refine` (object): Optional: the refine object a pending refine_model result carried, passed back unchanged.

### `look_at_model` (~174 tokens)

Look at a 3D model

FREE. See a 3D model instead of just linking to it. Renders the GLB from several angles and returns the frames as images you can actually look at, plus its geometry (triangles, materials, textures) and a plain reading of what those numbers mean. Use it to check your own work right after generating: is the subject complete, is the back finished, is anything melted or fused? Then regenerate naming the fault you saw. Works on any public https .glb, not only ones made here.

Input parameters:

- `glb_url` (string, required): Public https URL of the .glb to look at.
- `size` (integer): Pixel size of each square frame. Default 512.
- `views` (array): Angles to render. Default: three-quarter, front, side, back.

### `search_catalog` (~241 tokens)

Search the three.ws asset catalog

Free, no account needed. Search every ready-made asset three.ws publishes in one call: CC0 3D props and objects, rigged humanoid characters, and retargetable motion clips. Returns ids, titles, tags, licenses, and CDN urls, plus category facets to narrow the next query. Use this before generating anything: if the catalog already has the prop or character, dropping it in is instant and free where generating one is not. Feed a returned id to get_item_source for code that renders it on any site.

Input parameters:

- `category` (string): Exact category from a previous result's facets.
- `kind` (string): Narrow to one library: object (CC0 props), character (rigged humanoids), animation (motion clips).
- `limit` (integer)
- `offset` (integer): Page offset. Use next_offset from the previous response.
- `q` (string): Free text over titles, names, tags, and categories, e.g. "wooden chair", "office", "wave". Every word must match somewhere, so start broad.
- `tag` (string): Exact tag from a previous result's facets.

### `get_catalog_item` (~125 tokens)

Read one catalog item

Free, no account needed. Read one three.ws catalog item in full by its id from search_catalog (for example "object:adjustable_wrench"): metadata, license, CDN urls, the site links that preview or edit it, related items of the same kind, and which source frameworks get_item_source can emit for it. Use this after search_catalog to inspect one result before you embed it; for paste-ready code call get_item_source instead.

Input parameters:

- `id` (string, required): Catalog id as `<kind>:<name>` from search_catalog. A bare name is also accepted.

### `get_item_source` (~193 tokens)

Get paste-ready source for a catalog item

Free, no account needed. Returns working code that renders one three.ws catalog item on any site: the <agent-3d> web component (pinned to the exact published version with its SRI hash), the <model-viewer> tag our own browse grids use, plain three.js with GLTFLoader, or a React component. Motion clips return three.js code that parses the clip JSON and plays it. Omit `framework` to get the one that fits the item best; pass `all` to get every variant at once. Use this when you are ready to put a catalog item on a page or in an app; call get_catalog_item first if you still need its license or metadata.

Input parameters:

- `framework` (string): Which form to emit. Defaults to the recommended one for the item kind. "all" returns every applicable variant.
- `id` (string, required): Catalog id from search_catalog.

### `create_agent_persona` (~233 tokens)

Save a rigged model as a living, persistent agent body

Turn a generated GLB into a NAMED, persistent agent body: a "persona" the assistant reuses across turns and across sessions. The model is copied into durable storage so the body survives the source URL expiring, then registered under a stable persona_id. The returned view renders the LIVING body inline: it idles between turns, and persona_say makes it lip-sync and emote a reply. The persona_id is the handle: keep it and pass it to get_agent_persona or persona_say later to bring the exact same body back. Use this once a rigged model should become a recurring character the assistant speaks through. No sign-in required.

Input parameters:

- `glb_url` (string, required): Public https URL of the rigged GLB to embody (e.g. from forge_avatar).
- `name` (string, required): A display name for the persona, 1–80 characters.
- `source_prompt` (string): Optional: the prompt that generated this body, kept as provenance.
- `voice` (string): Optional voice name to speak with (used for audio-driven lip-sync when available).

### `get_agent_persona` (~84 tokens)

Reload a persona by id (continuity across sessions)

Bring back a previously saved persona by its persona_id: the SAME body and identity, in a fresh session. Returns the persona name, its model, the accumulated turn count, and the inline living-body view. Use this when a conversation starts and the user returns to a named agent.

Input parameters:

- `persona_id` (string, required): The persona_id returned by create_agent_persona.

### `persona_say` (~161 tokens)

Speak a reply through a persona: lip-sync + emotion + gesture

Make a persona PERFORM a reply: the body lip-syncs the text and shows the matching facial expression and body gesture. Pass the persona_id and the exact text the agent is saying this turn; the emotion is detected from the text automatically (or set it explicitly). The returned view animates the body for this turn: show it alongside the reply. Use this whenever the persona speaks, on every turn, with the persona_id from create_agent_persona or get_agent_persona.

Input parameters:

- `emotion` (string): Optional explicit emotion override; omit to auto-detect from the text.
- `persona_id` (string, required): The persona to speak through.
- `text` (string, required): The reply text the agent is saying this turn: drives lip-sync and emotion.

## Diagnostics

Captured diagnostic sections: TLS, DNSSEC, Authorisation, Transports. The full working is on the page: https://verifymcp.io/servers/nirholas-threews-grok/api-mcp-grok#diagnostics

## Score history

- 2026-10-23: 79
- 2026-10-22: 78
- 2026-10-21: 78
- 2026-10-20: 78
- 2026-10-19: 77
- 2026-10-18: 77
- 2026-10-17: 76
- 2026-10-16: 76
- 2026-10-15: 75
- 2026-10-14: 75
- 2026-10-13: 74
- 2026-10-12: 74

## Common questions

### What is the three.ws for Grok MCP server?

three.ws for Grok is an MCP server listed in the public MCP registry as io.github.nirholas/threews-grok. Free text/image to 3D for Grok Bot, plus your three.ws agents once signed in. Never spends. This page covers its hosted endpoint (https://three.ws/api/mcp-grok).

### Is the three.ws for Grok MCP server safe to use?

three.ws for Grok scores 79 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 three.ws for Grok MCP server expose?

three.ws for Grok exposes 15 tools: forge_free, text_to_avatar, mesh_forge, rig_mesh, forge_avatar, and 10 more. Their descriptions and schemas cost roughly 3,042 tokens of context every time the server is loaded.

### Does the three.ws for Grok MCP server require authentication?

No. We connected to three.ws for Grok without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

### Is the three.ws for Grok MCP server still maintained?

three.ws for Grok is still listed as active in the MCP registry. We last reached this channel on 23 October 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

## Links

- Remote endpoint: https://three.ws/api/mcp-grok
- Repository: https://github.com/nirholas/three.ws
- Website: https://three.ws/docs/grok
- Changelog RSS feed: https://verifymcp.io/servers/nirholas-threews-grok/api-mcp-grok.xml
- Changelog JSON feed: https://verifymcp.io/servers/nirholas-threews-grok/api-mcp-grok.json
- HTML version of this page: https://verifymcp.io/servers/nirholas-threews-grok/api-mcp-grok