# step2gcode (remote · www.step2gcode.com)

Turn STEP, DXF and SVG files into CNC milling and laser G-code for GRBL, LinuxCNC, Mach3 and more.

- Trust score: 73/100 (medium)
- Change this week: +4
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-11-21

## Components

- remote · `www.step2gcode.com`: 73/100, [markdown](https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp.md), [page](https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp)
- remote · `www.step2gcode.com`: 73/100 (this document), [markdown](https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp-test.md), [page](https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp-test)

## Channel facts

- Endpoint: `https://www.step2gcode.com/api/mcp/test`
- Transports: `streamable-http`
- Auth: `none`
- Version: `1.0.1`

## 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**: 80/100
  - The endpoint's TLS certificate is valid, in date, and uses a strong key.
  - No authorisation is required to call this server. Nothing on it is marked destructive, but we can't show that an anonymous caller can only read: full credit needs the server to list at least one tool and every tool to be verifiably read-only. We use a best-effort check of tool names and descriptions to tell read-only tools from tools that change things, so some genuinely read-only tools may not get full credit. Names that start with a read verb (get, list, read, search and similar) and accurate readOnlyHint annotations help. Meet that to earn full credit, or require authorisation.
  - HTTPS is enforced; there's no plaintext access path.
  - The HSTS (Strict-Transport-Security) header is present.
  - DNSSEC check failed: this domain isn't protected by DNSSEC.
- **Transport & Reachability**: 100/100
  - Verified streamable-http transport via a live MCP handshake.
- **Schema Quality & AI Usability**: 61/100
  - AI-judged instruction clarity (excellent).
  - Context-footprint check failed: tool/resource definitions use about 3321 tokens (~276/item across 12 items; 12 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 27/100
  - Stability observed for 8 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 97/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 88% of tool parameters carry a description.
  - Structured output schemas are declared (17% of tools); any adoption earns full credit.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 12 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 13 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 step2gcode MCP server?

step2gcode is a hosted endpoint at https://www.step2gcode.com/api/mcp/test, 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 com-step2gcode-step2gcode 'https://www.step2gcode.com/api/mcp/test'
```

### Cursor

```json
{
  "mcpServers": {
    "com-step2gcode-step2gcode": {
      "url": "https://www.step2gcode.com/api/mcp/test"
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "com-step2gcode-step2gcode": {
      "type": "http",
      "url": "https://www.step2gcode.com/api/mcp/test"
    }
  }
}
```

### Codex

```toml
[mcp_servers.com-step2gcode-step2gcode]
url = "https://www.step2gcode.com/api/mcp/test"
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "com-step2gcode-step2gcode": {
      "type": "remote",
      "url": "https://www.step2gcode.com/api/mcp/test",
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add com-step2gcode-step2gcode --url 'https://www.step2gcode.com/api/mcp/test' --transport streamable-http
```

### Hermes

```yaml
mcp_servers:
  com-step2gcode-step2gcode:
    url: "https://www.step2gcode.com/api/mcp/test"
```

### Netclaw

```json
{
  "McpServers": {
    "com-step2gcode-step2gcode": {
      "Transport": "http",
      "Url": "https://www.step2gcode.com/api/mcp/test"
    }
  }
}
```

### Vellum

```bash
assistant mcp add com-step2gcode-step2gcode -t streamable-http -u 'https://www.step2gcode.com/api/mcp/test'
```

### Other

```json
{
  "mcpServers": {
    "com-step2gcode-step2gcode": {
      "type": "http",
      "url": "https://www.step2gcode.com/api/mcp/test"
    }
  }
}
```

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 73, +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.

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

- [security] Tool “get_step_download_link” rewrote its description, which is the text the model reads

### 2026-11-18 (score 72, +1)

No change was recorded against any check on this day. Stability & Change Management went from 13 to 17. That category is still filling its 30-day observation window: 4 days of observed history at the previous scan, 5 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-16 (score 71, +1)

No change was recorded against any check on this day. Stability & Change Management went from 7 to 10. That category is still filling its 30-day observation window: 2 days of observed history at the previous scan, 3 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-11-15 (score 70, +1)

- [security] The server rewrote its instructions, which are the text every model session reads
- [functional] First check of Tool coverage: 17
- [functional] New tool “fetch”
- [functional] New tool “search”

### 2026-11-14 (score 69, 0)

- [functional improvement] Stability: unverified → 0.03

### 2026-11-13 (score 69)

First indexed and scored.

## MCP tools (12)

### `create_upload_link` (~146 tokens)

Create a short-lived link the user opens in their browser to upload a CAD file (STEP, DXF or SVG). Returns the link and an upload_id. Poll check_upload with the upload_id until it reports a file_id, then call generate_gcode. Free.

Before using this, check whether you can run shell commands. If you can, do not send the user to the browser: compress the file locally and pass it to upload_file_content (the compression is detected by itself). That works inside sandboxes without network access, which is the common case - a direct curl upload fails there with a proxy error. The response of this tool spells out all three routes in order.

### `check_upload` (~186 tokens)

Check whether the user finished the browser upload. Returns the file_id once ready. Free.

NEXT STEP - do not generate yet. Ask the user:
1\. Which material? Offer the presets (wood, hardwood, plywood, acrylic, aluminum, steel, brass). If theirs is not among them, pick the closest preset AND set spindle_speed, feed_rate, plunge_rate and step_down yourself from known cutting data for that material and cutter.
2\. Which cutter diameters do they actually own? Recommend one from that list: it must fit the smallest inner radius of the part, and a smaller cutter means longer machining time. If you cannot judge the geometry yet, say so and suggest generating once to see the dimensions.
3\. Is the part cut all the way through? Then recommend holding_tabs so it does not come loose on the last pass.

Input parameters:

- `upload_id` (string, required)

### `upload_file_content` (~498 tokens)

Upload a CAD file by writing its content out yourself. Use this only when a direct curl upload to /api/upload is not possible - that route costs you no output at all. If you must use this tool, ALWAYS compress first and pass the output as content:
  (xz -9e -c model.step 2>/dev/null || gzip -9c model.step) | base64 | tr -d '\n'
xz writes about a quarter less than gzip; the command falls back to gzip where xz is missing (macOS). xz, bzip2 and gzip are recognised by themselves - no extra argument.

Do not paste the raw file. You have to write out every character you pass here, and a typical STEP solid takes you three minutes raw against twenty seconds compressed - the user watches that happen. Compression shrinks it about fivefold, which brings files up to roughly 80 KB within reach.

NO SHELL AVAILABLE? Then do not use this tool for anything above about 10 KB - you would have to type the whole file out and the user waits minutes for it. Call create_upload_link instead: the user drops the file in a browser and it is done in five seconds. Same if the file is still too large after compressing. Free.

NEXT STEP - do not generate yet. Ask the user:
1\. Which material? Offer the presets (wood, hardwood, plywood, acrylic, aluminum, steel, brass). If theirs is not among them, pick the closest preset AND set spindle_speed, feed_rate, plunge_rate and step_down yourself from known cutting data for that material and cutter.
2\. Which cutter diameters do they actually own? Recommend one from that list: it must fit the smallest inner radius of the part, and a smaller cutter means longer machining time. If you cannot judge the geometry yet, say so and suggest generating once to see the dimensions.
3\. Is the part cut all the way through? Then recommend holding_tabs so it does not come loose on the last pass.

Input parameters:

- `content` (string, required): Output of: (xz -9e -c model.step 2>/dev/null || gzip -9c model.step) | base64 | tr -d '\n' - the compressed file, base64 encoded (xz, bzip2 or gzip, detected by itself).
- `filename` (string, required): File name including extension, e.g. bracket.step

### `generate_gcode` (~874 tokens)

Start G-code generation for an uploaded file and wait for the result. Most jobs finish within the wait, so the answer usually already carries the part dimensions and hole diameters - no polling needed. Only if it says the job is still running, call get_job_status once after a few seconds. Free - only the download is charged.

BEFORE CALLING THIS, ask the user about the settings you cannot infer from the file. Two of them change the result substantially and are wrong by default more often than not:
\- material: drives spindle speed, feed rates and depth per pass. Aluminium cut with wood settings breaks cutters.
\- holding_tabs: without them a cut-out part comes loose on the last pass and can be thrown off the machine. Recommend them for any part cut all the way through.
Also worth asking when the geometry suggests it: pockets and holes (STEP files carry that depth information), and whether the job is milling or laser work.

IMAGES (PNG, JPG, WebP ...) are always laser raster engraving. Ask for the engraving width in mm, photo or logo (image_mode), the controller (GRBL or Marlin) and its S maximum (laser_max_s), and tell the user to test power and speed on a scrap piece first.

Input parameters:

- `bit_diameter` (number): Cutter diameter in mm, e.g. 6. Leave it out and step2gcode picks the largest whole millimetre that still fits through the smallest hole in the part.
- `controller` (string): Laser G-code dialect. GRBL (default) or Marlin (3D printers with a laser module, MPCNC/LowRider and other Marlin machines).
- `depth` (number): Cutting depth in mm; omit to take it from the model geometry (STEP)
- `feed_rate` (number): Horizontal feed in mm/min. Same rule: only for materials outside the presets.
- `file_id` (string, required): file_id from check_upload or upload_file_content
- `holding_tabs` (boolean): Leave small bridges so a part cut all the way through stays in the stock. Strongly recommended for through cuts.
- `holes` (boolean): Drill or mill the holes found in the geometry
- `image_mode` (string): Images only. Line by line (raster): dither for photos (default), threshold for logos and text, grayscale for power per dot. Along lines (vector, much faster): outline traces the edges of dark areas,…
- `image_width_mm` (number): Images only: engraving width in mm; the height follows from the aspect ratio. Default 100.
- `inner_contours` (boolean): Cut inner contours and cut-outs
- `laser` (boolean): Laser job instead of milling
- `laser_max_s` (number): Highest S value of the machine: $30 on GRBL (usually 1000), 255 on most Marlin machines. Ask the user - a wrong value burns at the wrong power.
- `laser_power` (number): Laser power in % (laser jobs). For images the default is 30 %.
- `laser_speed` (number): Laser speed in mm/min (laser jobs). For images the default is 2000.
- `material` (string): Workpiece material. Sets spindle speed, feed rates and depth per pass. Ask the user - the default assumes softwood.
- `origin_xy` (string): Workpiece origin in XY. Default TL.
- `origin_z` (string): Z zero on the stock surface (top) or on the spoilboard. Default top.
- `plunge_rate` (number): Plunge feed in mm/min, for materials outside the presets.
- `pockets` (boolean): Mill pockets found in the geometry (STEP)
- `ramp_entry` (boolean): Enter the material on a ramp instead of plunging straight down
- `spindle_speed` (number): Spindle speed in rpm. Only set this when the material is not covered by the presets - otherwise the preset value is used.
- `step_down` (number): Depth per pass in mm. Rule of thumb: up to 1x cutter diameter in softwood, far less in metal.

### `get_job_status` (~100 tokens)

Check a generation job. While it runs you get the percentage done, how long it has been running and how much time is left before the generator gives up on its own (180 s). When it fails you get the generator's own reason - pass that on to the user instead of retrying blindly. This is the ONLY correct way to wait for a job; calling generate_gcode again never helps. Free.

Input parameters:

- `job_id` (string, required)

### `preview_toolpath` (~55 tokens)

Show the finished tool path as an image (top view and isometric view, cutting moves and rapids in different colours) so the user can sanity-check the job before paying. Free.

Input parameters:

- `job_id` (string, required)

### `get_download_link` (~130 tokens)

Return the download link for the finished G-code. THIS is the billed step: one credit per model not paid for yet. Re-downloading the same job or a model already paid for is free.

Called without confirm, this tool charges nothing and only reports what the download would cost. Show that to the user, wait for their answer, and only then call again with confirm=true. Never set confirm=true on your own initiative - it spends the user's money.

Input parameters:

- `confirm` (boolean): Set to true only after the user has explicitly agreed to spend the credit.
- `job_id` (string, required)

### `convert_mesh_to_step` (~253 tokens)

Turn an uploaded STL or 3MF mesh back into a STEP solid with exact surfaces - planes, cylinders, cones and fillets instead of triangles, intersected into real edges. Free; only the download of the finished file is charged.

Use this when the user has a mesh and needs CAD geometry: a hole becomes a cylinder with a diameter again, a chamfer a cone with an angle, so the part can be dimensioned, modified and machined. Typical case: the CAD source is gone and the mesh is all that is left.

The answer states the largest measured deviation between the built solid and the mesh, in millimetres - always pass that number on to the user, it is what tells them whether the result is accurate enough for their part. Parts made of freeform surfaces and raw 3D scans are rejected with a reason instead of being badly reconstructed.

Upload the mesh the same way as any other file (curl to /api/upload, upload_file_content or create_upload_link), then pass the file_id here. The resulting STEP file can go straight into generate_gcode afterwards.

Input parameters:

- `file_id` (string, required): file_id of an uploaded .stl or .3mf mesh

### `get_step_download_link` (~170 tokens)

Return the download link for a STEP file built by convert_mesh_to_step. While stl2step is in beta this is free and the answer says so; otherwise it is the billed step: one credit per model not paid for yet. The DXF contour that comes with a 2.5D part belongs to the same part and costs nothing extra.

Called without confirm, this tool charges nothing and only reports what the download would cost. Show that to the user, wait for their answer, and only then call again with confirm=true. Never set confirm=true on your own initiative - it spends the user's money.

Input parameters:

- `confirm` (boolean): Set to true only after the user has explicitly agreed to spend the credit.
- `step_id` (string, required): step_id from convert_mesh_to_step

### `search` (~97 tokens)

Search step2gcode's help and knowledge base: supported files (STEP, DXF, SVG, STL, images, Gerber), machines and post-processors, settings, laser, 3D milling, pricing and credits. Returns pages with id, title and url; read one with fetch. Free, no file needed.

Input parameters:

- `query` (string, required): What to look for, e.g. 'DXF holes smaller than the cutter'

Output parameters:

- `results` (array)

### `fetch` (~37 tokens)

Read the full text of one page found with search, by its id. Free.

Input parameters:

- `id` (string, required): The id from a search result

Output parameters:

- `id` (string)
- `text` (string)
- `title` (string)
- `url` (string)

### `get_credit_balance` (~27 tokens)

Show credits left on the access token and how many models are already paid for. Free.

## Diagnostics

Captured diagnostic sections: TLS, DNSSEC, Authorisation, Transports. The full working is on the page: https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp-test#diagnostics

## Score history

- 2026-11-21: 73
- 2026-11-20: 72
- 2026-11-19: 72
- 2026-11-18: 72
- 2026-11-17: 71
- 2026-11-16: 71
- 2026-11-15: 70
- 2026-11-14: 69
- 2026-11-13: 69

## Common questions

### What is the step2gcode MCP server?

step2gcode is an MCP server listed in the public MCP registry as com.step2gcode/step2gcode. Turn STEP, DXF and SVG files into CNC milling and laser G-code for GRBL, LinuxCNC, Mach3 and more. This page covers its hosted endpoint (https://www.step2gcode.com/api/mcp/test).

### Is the step2gcode MCP server safe to use?

step2gcode scores 73 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 step2gcode MCP server expose?

step2gcode exposes 12 tools: create_upload_link, check_upload, upload_file_content, generate_gcode, get_job_status, and 7 more. Their descriptions and schemas cost roughly 2,573 tokens of context every time the server is loaded.

### Does the step2gcode MCP server require authentication?

No. We connected to step2gcode without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

### Is the step2gcode MCP server still maintained?

step2gcode 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 step2gcode MCP server?

VerifyMCP ran 13 checks on step2gcode, 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://www.step2gcode.com/api/mcp/test
- Website: https://www.step2gcode.com/en/mcp
- Changelog RSS feed: https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp-test.xml
- Changelog JSON feed: https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp-test/changelog.json
- Trust report (JSON): https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp-test.json
- HTML version of this page: https://verifymcp.io/servers/com-step2gcode-step2gcode/api-mcp-test
