Agent Tool

Scan and Mesh Workshop

Name: Scan and Mesh Workshop
Brand: Apoth3osis
SKU: 6ae3e6dc6b8e2f458be7d95c
Price: 0.03 USD
Availability: InStock

Available ActionsEach successful request consumes credits as outlined below.

inspect_geometry^2crcreate_geometry^2crprocess_point_cloud^3crprocess_mesh^3crreconstruct_surface^8crregister_geometry^5crcompare_geometry^5crmeasure_geometry^2crsegment_geometry^5crrender_geometry^2crexport_geometry^2crrun_script^5crget_tasklist_taskscancel_task

Details

Turn raw 3D scans, point clouds and meshes into clean, measured, ready-to-use geometry in one place. Load common mesh, point cloud and STEP files and get an immediate quality report. Clean noisy scans, downsample, rebuild surfaces, or fuse depth camera frames into a mesh. Repair meshes by welding, removing defects and filling holes so they follow the surrounding curvature, then simplify, smooth or combine them with watertight booleans. Align scans automatically, compare a scan with its design to see deviation maps and the share within tolerance, and measure sizes, cross sections, flat faces and wall thickness. Split scenes into planes, clusters or AI-labeled parts such as walls, floors, furniture, roads and vehicles. Render preview images and export to standard formats in the units you need.

Use Cases

Clean a noisy 3D scan and turn it into a watertight mesh, Compare a scanned part against its CAD design, Repair a broken STL before 3D printing, Align several scans of the same object, Measure wall thickness and sections from a scan, Label walls, floors and furniture in a room scan, Convert meshes between formats and units

Dynamic MCP Setup

Connect once through AgentPMT Dynamic MCP, then use approved tools from the same agent connection.

30 Second Setup

STDIO connector for Claude Code, Codex, Cursor, Zed, and other LLMs that require STDIO or custom connections.

npm install -g @agentpmt/mcp-routeragentpmt-setup

Hosted Streamable HTTPS

MCP endpoint for browser-based apps like ChatGPT, Claude, Grok, or any time you want a streamable connection with no local install.

https://api.agentpmt.com/mcp

Config Example

Use the hosted endpoint directly in clients that support remote MCP. Store your Bearer token in the client config or secret field.

Full connection guide

{
  "mcpServers": {
    "agentpmt": {
      "type": "streamable-http",
      "url": "https://api.agentpmt.com/mcp",
      "headers": {
        "Authorization": "Bearer <AGENTPMT_BEARER_TOKEN>",
        "x-instance-metadata": "{\"client\":\"generic-mcp\",\"platform\":\"remote\"}"
      }
    }
  }
}

Need client videos, organization controls, audit details, and the full feature overview?

More About Dynamic MCP

Actions(15)

inspect_geometry^2cr8 params

Load a mesh, point cloud or STEP file, or re-check an existing geometry. Returns a geometry_id and a quality report: counts, size, watertightness, self-intersections, holes, pieces, normals, colors, point spacing, suggested next steps and a preview image.

file_idstring

File Manager ID of a ply, obj, stl, off, glb, gltf, pcd, xyz, xyzn, xyzrgb, pts, step or stp file. Give file_id or geometry_id.

geometry_idstring

Existing geometry to re-check instead of loading a file.

unitsstring

Units of the file. Required for stl, obj, ply, off, pcd, xyz, xyzn, xyzrgb and pts, which record none. glb and gltf are meters and STEP declares its own; giving units converts to them.

Values:

mmcmmin

up_axisstring

Axis that points up in the file. Default y for glb and gltf, z otherwise. Stored z-up.

Values:

xyz

load_asstring

auto loads files with faces as meshes. Default auto.

Values:

automeshpoint_cloud

step_tolerancenumber

STEP only: maximum mesh deviation from the exact surface in file units. Default 0.1 percent of the part size.

Range: 0.0001 - ...

previewboolean

Return a preview image of the result. Default true.

namestring

Readable name used in file names.

create_geometry^2cr12 params(1 required)

Create a mesh primitive, 3D text or an extruded outline, ready to process, combine with booleans, measure or export.

shaperequiredstring

Shape to create.

Values:

boxspherecylinderconetorusarrowtetrahedronoctahedronicosahedrontextextruded_outline

widthnumber

Box size along x.

Range: 0 - ...

depthnumber

Box size along y, or text thickness.

Range: 0 - ...

heightnumber

Box, cylinder, cone or extrusion height, arrow length, or text height.

Range: 0 - ...

radiusnumber

Sphere, cylinder, cone, torus or polyhedron radius.

Range: 0 - ...

tube_radiusnumber

Torus tube radius.

Range: 0 - ...

resolutioninteger

Segments around round shapes. Default 64.

Range: 3 - 1024

textstring

Text for the text shape.

outlinearray

Closed outline for extruded_outline as [x, y] points in order.

Array of: array

centerarray

Center position. Default the origin.

Array of: number

unitsstring

Units for all sizes. Default mm.

Values:

mmcmmin

namestring

Readable name used in file names.

process_point_cloud^3cr4 params(2 required)

Clean and prepare a point cloud with ordered steps: downsampling, outlier removal, normals, cropping, transforms, merging and keeping segment labels. Returns a new geometry and a per-step report.

geometry_idrequiredstring

geometry_id returned by inspect_geometry or an earlier action.

stepsrequiredarray

Steps to run in order.

Array of: object

previewboolean

Return a preview image of the result. Default true.

namestring

Readable name used in file names.

process_mesh^3cr4 params(2 required)

Repair and modify a mesh with ordered steps: welding, cleanup, hole filling that follows the surrounding curvature, self-intersection repair, simplification, smoothing, subdivision, watertight booleans, piece filtering, transforms, ambient occlusion and point sampling. Every result is checked for watertightness.

geometry_idrequiredstring

geometry_id returned by inspect_geometry or an earlier action.

stepsrequiredarray

Steps to run in order.

Array of: object

previewboolean

Return a preview image of the result. Default true.

namestring

Readable name used in file names.

reconstruct_surface^8cr12 params

Build a mesh from a point cloud, or fuse RGB-D camera frames into a mesh. auto picks the method and settings from the data.

geometry_idstring

Point cloud to reconstruct. Not used with rgbd_fusion.

methodstring

Default auto. poisson gives a closed watertight surface, ball_pivoting keeps the original points, alpha_shape suits sparse data, rgbd_fusion integrates camera frames.

Values:

autopoissonball_pivotingalpha_shapergbd_fusion

detailstring

Surface detail for poisson and rgbd_fusion. Default medium.

Values:

lowmediumhigh

trim_low_densitynumber

poisson: remove this fraction of the least supported surface, which opens holes. Default 0.

Range: 0 - 0.5

radiusnumber

ball_pivoting radius or alpha_shape alpha in geometry units. Default from point spacing.

Range: 0 - ...

framesarray

rgbd_fusion: color and depth image pairs in capture order.

Array of: object

cameraobject

rgbd_fusion: pinhole intrinsics in pixels.

Properties:

width*(integer)- Image width.

height*(integer)- Image height.

fx*(number)- Focal length x.

fy*(number)- Focal length y.

cx*(number)- Principal point x.

cy*(number)- Principal point y.

depth_scalenumber

rgbd_fusion: depth units per meter. Default 1000.

Range: 0 - ...

depth_maxnumber

rgbd_fusion: ignore depth beyond this many meters. Default 3.

Range: 0 - ...

unitsstring

rgbd_fusion: units for the fused mesh. Default m.

Values:

mmcmmin

previewboolean

Return a preview image of the result. Default true.

namestring

Readable name used in file names.

register_geometry^5cr7 params(2 required)

Align scans to a target automatically (feature matching, then ICP). Returns the transform, fit quality, aligned geometries, an overlay image, and a warning when a symmetric part also fits flipped.

source_geometry_idrequiredstring

Geometry to move.

target_geometry_idrequiredstring

Geometry to align to.

extra_source_geometry_idsarray

More scans, each aligned to the target the same way.

Array of: string

methodstring

Default auto: global feature matching, then point-to-plane ICP. ICP options refine from initial_transform.

Values:

autoicp_point_to_pointicp_point_to_planeicp_coloredicp_generalized

initial_transformarray

Transform as 16 numbers, row by row, in the geometry units.

Array of: number

max_distancenumber

Largest point pair distance to trust, in target units. Default from point spacing.

Range: 0 - ...

applyboolean

Also return the aligned sources as new geometries. Default true.

compare_geometry^5cr6 params(2 required)

Compare a scan or mesh with a reference: deviation statistics, percent within tolerance, signed inside or outside deviation for closed references, Chamfer, Hausdorff and F-score, plus a colored deviation map.

geometry_idrequiredstring

Measured geometry, such as a scan.

reference_geometry_idrequiredstring

Reference geometry, such as the design mesh.

tolerancenumber

Deviation counted as in tolerance, in reference units. Default 0.1 mm.

Range: 0 - ...

align_firstboolean

Align automatically before comparing. Default false.

f_score_radiusnumber

Distance for the F-score. Default tolerance.

Range: 0 - ...

namestring

Readable name used in file names.

measure_geometry^2cr2 params(1 required)

Measure a geometry: size and oriented box, area and volume, cross sections, flat faces, wall thickness between parallel faces, distance to another geometry, and convex hull.

geometry_idrequiredstring

geometry_id returned by inspect_geometry or an earlier action.

measurementsarray

Measurements to take. Default overview.

Array of: object

segment_geometry^5cr10 params(2 required)

Split a point cloud or mesh into labeled parts: planes, flat patches, clusters, connected pieces, AI scene labels (wall, floor, table, road, building, car and more) or AI 3D object boxes. Returns a labeled geometry for keep_labels, a per-segment summary and a colored preview.

geometry_idrequiredstring

geometry_id returned by inspect_geometry or an earlier action.

methodrequiredstring

planes, planar_patches, clusters, connected_components, semantic (AI labels per point) or objects (AI 3D boxes).

Values:

planesplanar_patchesclustersconnected_componentssemanticobjects

scenestring

How the scan was captured; picks the AI model. indoor_room: rooms and buildings. outdoor_street: street-level mobile or terrestrial scans. vehicle_lidar: one sweep from a car-mounted spinning LiDAR with the sensor at the origin (needed for objects).

Values:

indoor_roomoutdoor_streetvehicle_lidar

modelstring

Override the model chosen from scene.

Values:

randlanet_s3diskpconv_s3dispointtransformer_s3dissparseconvunet_scannetrandlanet_semantickittikpconv_semantickittirandlanet_toronto3dkpconv_toronto3drandlanet_semantic3drandlanet_parislille3dkpconv_parislille3dpointpillars_kittipointpillars_waymopointrcnn_kitti

max_segmentsinteger

Maximum planes, patches, clusters or pieces. Default 10.

Range: 1 - 200

distance_thresholdnumber

planes: distance to count as on a plane. clusters: neighbor distance. Default from point spacing.

Range: 0 - ...

min_pointsinteger

Smallest segment to keep. Default 50.

Range: 1 - ...

min_confidencenumber

objects: lowest box confidence to return. Default 0.5.

Range: 0 - 1

previewboolean

Return a preview image of the result. Default true.

namestring

Readable name used in file names.

render_geometry^2cr7 params(1 required)

Render PNG images of one or more geometries together, colored by their own colors (deviation maps, labels), height, normals or a uniform color.

geometry_idsrequiredarray

Geometries to draw in one scene.

Array of: string

viewsarray

Views to render. Default iso.

Array of: string

directionarray

Custom view direction from the geometry toward the camera.

Array of: number

color_bystring

Default natural: stored colors such as deviation maps and labels.

Values:

naturalheightnormalsuniform

point_sizenumber

Point size in pixels for point clouds. Default 2.

Range: 1 - 10

width_pxinteger

Image width. Default 1024.

Range: 256 - 2048

height_pxinteger

Image height. Default 768.

Range: 256 - 2048

export_geometry^2cr5 params(2 required)

Export a geometry to a mesh or point cloud file in the units you choose. glb and gltf are always meters, y-up, as the standard requires.

geometry_idrequiredstring

geometry_id returned by inspect_geometry or an earlier action.

formatrequiredstring

Output format. stl, obj, off, glb and gltf need a mesh.

Values:

plyobjstloffglbgltfpcdxyzxyznxyzrgbpts

unitsstring

Units for the exported coordinates. Default the geometry units.

Values:

mmcmmin

binaryboolean

Write binary where the format allows. Default true.

filenamestring

Base file name without folder or extension.

run_script^5cr2 params(1 required)

Run custom Open3D Python on your geometries for anything the other actions do not cover. Inputs are variables; publish(name, geometry) returns new geometries and report({...}) returns values.

scriptrequiredstring

Python with o3d (open3d) and np (numpy) preloaded; scipy and manifold3d importable. No network access.

inputsarray

Geometries loaded as Open3D objects under these variable names.

Array of: object

get_task2 params(1 required)

Check work that returned status processing. Completed tasks return the original action result in outputs[0]. Free.

task_idrequiredstring

task_id from a processing response.

wait_secondsinteger

Seconds to wait for completion (0-55); use 55 when polling. Default 0.

Range: 0 - 55

list_tasks1 param

List your recent tasks with their status and the geometries they produced, newest first. Free.

limitinteger

Maximum results. Default 20.

Range: 1 - 100

cancel_task1 param(1 required)

Cancel work that is still processing. Free.

task_idrequiredstring

task_id from a processing response.

curl -X POST "https://api.agentpmt.com/products/purchase" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ********" \
  -d '{
    "product_id": "6ae3e6dc6b8e2f458be7d95c",
    "parameters": {
      "action": "inspect_geometry"
    }
  }'

import requests
import json

url = "https://api.agentpmt.com/products/purchase"

headers = {
    "Content-Type": "application/json",
    "Authorization": "Bearer ********"
}

data = {
    "product_id": "6ae3e6dc6b8e2f458be7d95c",
    "parameters": {
        "action": "inspect_geometry"
    }
}

response = requests.post(url, headers=headers, json=data)
print(response.status_code)
print(response.json())

const url = "https://api.agentpmt.com/products/purchase";

const headers = {
  "Content-Type": "application/json",
  "Authorization": "Bearer ********"
};

const data = {
  product_id: "6ae3e6dc6b8e2f458be7d95c",
  parameters: {
    "action": "inspect_geometry"
  }
};

fetch(url, {
  method: "POST",
  headers,
  body: JSON.stringify(data)
})
  .then(response => response.json())
  .then(data => console.log(data))
  .catch(error => console.error("Error:", error));

const axios = require('axios');

const url = "https://api.agentpmt.com/products/purchase";

const headers = {
  "Content-Type": "application/json",
  "Authorization": "Bearer ********"
};

const data = {
  product_id: "6ae3e6dc6b8e2f458be7d95c",
  parameters: {
    "action": "inspect_geometry"
  }
};

axios.post(url, data, { headers })
  .then(response => {
    console.log(response.status);
    console.log(response.data);
  })
  .catch(error => {
    console.error("Error:", error.message);
  });

Login to view your API and budget keys. The example above uses placeholder values. Sign in to see personalized code with your bearer token.

Usage Instructions

Usage guidance provided directly by the developer for this product.

Scan and Mesh Workshop

Load scans, point clouds and meshes, then clean, repair, reconstruct, align, compare, measure, segment, render and export them; every result comes with a quality report and a preview image.

Workflow

inspect_geometry with a file_id (and units for formats without units) returns a geometry_id, a report and next_steps.
Pass that geometry_id to other actions. Each action that produces geometry returns new geometry_ids; the inputs are never changed.
export_geometry writes a standard file when you are done.

Units are set once when loading; all later sizes and settings are in the geometry's units. Any action that runs the geometry engine returns its result directly when it finishes within about 20 seconds. Longer work returns status: "processing" with a task_id; call get_task with wait_seconds: 55 until completed (the result is in outputs[0]) or failed. Tasks are saved before work starts, so they survive server restarts (an interrupted job restarts automatically from the beginning) and wait in a queue when every engine is busy; you never need to resubmit a processing task.

Actions

`get_instructions`

Returns this documentation.

`inspect_geometry`

Required: file_id or geometry_id. Optional: units (mm, cm, m, in; required for stl, obj, ply, off, pcd, xyz, xyzn, xyzrgb, pts), up_axis (x, y, z; default y for glb/gltf, z otherwise), load_as (auto, mesh, point_cloud), step_tolerance, preview, name. Formats: ply, obj, stl, off, glb, gltf, pcd, xyz, xyzn, xyzrgb, pts, step, stp. Points with missing (NaN or infinite) coordinates are dropped on load and counted in load_notes.removed_non_finite_points.

{"action":"inspect_geometry","file_id":"FILE_ID","units":"mm"}

`create_geometry`

Required: shape (box, sphere, cylinder, cone, torus, arrow, tetrahedron, octahedron, icosahedron, text, extruded_outline). Sizes: width, depth, height, radius, tube_radius, resolution, text, outline ([[x, y], ...]), center, units, name.

{"action":"create_geometry","shape":"extruded_outline","outline":[[0,0],[30,0],[30,10],[10,10],[10,30],[0,30]],"height":5}

`process_point_cloud`

Required: geometry_id, steps (run in order). Operations: voxel_downsample (size), uniform_downsample (every_k), random_downsample (ratio), farthest_point_downsample (count), remove_statistical_outliers (count neighbors, ratio), remove_radius_outliers (size, count), remove_non_finite, estimate_normals (size, count; orients them too), orient_normals (vector, or automatic when omitted: consistent across the surface, along the plane normal for flat clouds; the method is in settings), crop_box (min, max), keep_labels (labels from segment_geometry), hidden_point_removal (vector), transform (matrix), translate (vector), rotate (vector axis, angle_deg), scale (factor), paint (color), merge (geometry_ids). Omitted settings are derived from the data and reported in steps[].settings.

{"action":"process_point_cloud","geometry_id":"GEOMETRY_ID","steps":[{"operation":"remove_statistical_outliers"},{"operation":"voxel_downsample"},{"operation":"estimate_normals"}]}

`process_mesh`

Required: geometry_id, steps. Operations: weld_vertices (size), remove_duplicates, remove_degenerate, remove_non_manifold_edges, orient_triangles (consistent winding; turns an inside-out closed mesh right side out), fill_holes (mode smooth or flat, max_hole_edges), repair_self_intersections, simplify (count target triangles), cluster_simplify (size), smooth (mode laplacian or taubin, count), subdivide (mode loop or midpoint, count), boolean (mode union, difference or intersection, geometry_ids), keep_largest_component, remove_small_components (count), crop_box, transform, translate, rotate, scale, compute_normals, bake_ambient_occlusion (count rays; stored as vertex colors), sample_points (count, mode uniform or poisson; returns a point cloud), convex_hull, paint.

Smooth hole filling continues the surrounding curvature; large patches are flagged as estimates. Booleans need closed meshes and always return watertight results.

{"action":"process_mesh","geometry_id":"GEOMETRY_ID","steps":[{"operation":"weld_vertices"},{"operation":"fill_holes","mode":"smooth"},{"operation":"keep_largest_component"}]}

{"action":"process_mesh","geometry_id":"PART_ID","steps":[{"operation":"boolean","mode":"difference","geometry_ids":["TOOL_ID"]}]}

`reconstruct_surface`

Required: geometry_id (point cloud), or method: "rgbd_fusion" with frames and camera. Optional: method (auto, poisson, ball_pivoting, alpha_shape, rgbd_fusion), detail (low, medium, high), trim_low_density (poisson; opens holes), radius, depth_scale, depth_max, units, preview, name. Poisson returns a closed, watertight surface.

{"action":"reconstruct_surface","geometry_id":"GEOMETRY_ID","method":"poisson","detail":"high"}

{"action":"reconstruct_surface","method":"rgbd_fusion","frames":[{"color_file_id":"C1","depth_file_id":"D1"},{"color_file_id":"C2","depth_file_id":"D2"}],"camera":{"width":640,"height":480,"fx":525,"fy":525,"cx":319.5,"cy":239.5}}

`register_geometry`

Required: source_geometry_id, target_geometry_id. Optional: extra_source_geometry_ids, method (auto, icp_point_to_point, icp_point_to_plane, icp_colored, icp_generalized), initial_transform (16 numbers, row by row), max_distance, apply (default true). Auto needs no starting pose. Returns transform, fitness, inlier_rmse, aligned geometries, an overlay image, and a warning when a symmetric part also fits flipped.

{"action":"register_geometry","source_geometry_id":"SCAN_ID","target_geometry_id":"CAD_ID"}

`compare_geometry`

Required: geometry_id (measured), reference_geometry_id. Optional: tolerance (default 0.1 mm), align_first, f_score_radius, name. Returns mean, rms, p95, max, within_tolerance_percent, signed inside/outside deviation for closed references, Chamfer, Hausdorff and F-score, plus a colored deviation map (green within tolerance, red extra material, blue missing). F-score also measures coverage, so it is low when the scan is sparser than the radius.

{"action":"compare_geometry","geometry_id":"SCAN_ID","reference_geometry_id":"CAD_ID","tolerance":0.15,"align_first":true}

`measure_geometry`

Required: geometry_id. Optional: measurements (default overview). Kinds: overview, section (plane, offset; meshes; returns enclosed_area with holes subtracted and separate pieces added, region_count, hole_count, outline_length and each loop's area), planes (count, distance_threshold), thickness (two best-supported parallel faces), distance (other_geometry_id), convex_hull.

{"action":"measure_geometry","geometry_id":"GEOMETRY_ID","measurements":[{"kind":"thickness"},{"kind":"section","plane":"Z","offset":2.5}]}

`segment_geometry`

Required: geometry_id, method (planes, planar_patches, clusters, connected_components, semantic, objects). Optional: scene (indoor_room, outdoor_street, vehicle_lidar), model, max_segments, distance_threshold, min_points, min_confidence, preview, name. AI methods need scene or model; objects needs a vehicle_lidar sweep with the sensor at the origin. AI models expect real scans: colors improve indoor and street labels. Returns a labeled geometry for keep_labels and a per-segment summary.

{"action":"segment_geometry","geometry_id":"ROOM_SCAN_ID","method":"semantic","scene":"indoor_room"}

{"action":"segment_geometry","geometry_id":"GEOMETRY_ID","method":"planes","max_segments":4}

`render_geometry`

Required: geometry_ids (1-10, drawn together). Optional: views (iso, top, bottom, front, back, left, right), direction, color_by (natural, height, normals, uniform), point_size, width_px, height_px.

{"action":"render_geometry","geometry_ids":["SCAN_ID","CAD_ID"],"views":["iso","top"]}

`export_geometry`

Required: geometry_id, format (ply, obj, stl, off, glb, gltf, pcd, xyz, xyzn, xyzrgb, pts). Optional: units, binary, filename. stl, obj, off, glb and gltf need a mesh; glb and gltf are always meters, y-up.

{"action":"export_geometry","geometry_id":"GEOMETRY_ID","format":"stl","units":"mm"}

`run_script`

Required: script. Optional: inputs (array of {name, geometry_id}). o3d and np are preloaded; inputs are Open3D TriangleMesh or PointCloud objects. Call publish(name, geometry) to return geometry and report({...}) to return values. No network access.

{"action":"run_script","script":"hull, _ = part.compute_convex_hull()\npublish('hull', hull)\nreport({'hull_volume': hull.get_volume()})","inputs":[{"name":"part","geometry_id":"GEOMETRY_ID"}]}

`get_task`

Required: task_id. Optional: wait_seconds (0-55; returns as soon as the task finishes). stage shows progress: queued, loading, the running operation, saving. restarts appears if the job was interrupted and restarted.

{"action":"get_task","task_id":"TASK_ID","wait_seconds":55}

`list_tasks`

Optional: limit (1-100, default 20).

{"action":"list_tasks"}

`cancel_task`

Required: task_id. Stops a queued or running job; returns cancelled: true once it has stopped. A job that is already saving its results completes instead (cancelled: false). A cancelled job creates nothing.

{"action":"cancel_task","task_id":"TASK_ID"}

Response

Actions that create geometry return geometries, each with geometry_id, kind, units, report (counts, size, watertight, holes, pieces, self-intersections, volume, point spacing, labels), next_steps and previews. Exports return files; renders return images. Every file has file_id and signed_url. Generated files are kept for 7 days.

Errors

MESH_UNITS_REQUIRED, MESH_UNSUPPORTED_FORMAT, MESH_LOAD_FAILED, MESH_EMPTY_GEOMETRY, MESH_NOT_A_MESH, MESH_NOT_A_POINT_CLOUD, MESH_INVALID_PARAMETER, MESH_STEP_FAILED, MESH_BOOLEAN_FAILED, MESH_RECONSTRUCTION_FAILED, MESH_REGISTRATION_FAILED, MESH_MODEL_UNAVAILABLE, MESH_MODEL_INPUT_MISMATCH, MESH_MODEL_FAILED, MESH_SCRIPT_ERROR, MESH_WORKSHOP_GEOMETRY_NOT_FOUND, MESH_WORKSHOP_GEOMETRY_EXPIRED, MESH_WORKSHOP_UNSUPPORTED_FILE, GEOMETRY_TIMEOUT, GEOMETRY_MEMORY_LIMIT, GEOMETRY_ENGINE_CRASH, GEOMETRY_INPUT_TOO_LARGE, GEOMETRY_INPUT_EXPIRED, GEOMETRY_WORKER_BUSY, GEOMETRY_WORKER_UNAVAILABLE, GEOMETRY_TASK_NOT_FOUND, GEOMETRY_TASK_INTERRUPTED. Each error includes a resolution with the next step.

Frequently Asked Questions

How do I connect this tool to an external agent?

You can install the local MCP server by opening a terminal and running:

Install commands

npm install -g @agentpmt/mcp-router
agentpmt-setup

This will connect you to local agents like Claude Code, Windsurf, Grok Build, Cursor, etc.

Alternatively you can connect to the hosted version with this config block, no installation required:

Hosted MCP config

{
  "mcpServers": {
    "agentpmt": {
      "type": "streamable-http",
      "url": "https://api.agentpmt.com/mcp",
      "headers": {
        "Authorization": "Bearer <AGENTPMT_BEARER_TOKEN>",
        "x-instance-metadata": "{\"client\":\"generic-mcp\",\"platform\":\"remote\"}"
      }
    }
  }
}

View MCP Connection Instructions for more details.

How does an external agent use this tool?

After the external agent is connected to an Agent Group that can use this tool, paste this prompt into the agent:

Agent prompt

Use the AgentPMT-Tool-Search-and-Execution tool. First call action 'get_instructions' so you know how to use the tool search interface. Then call action 'get_schema' with tool_id 6ae3e6dc6b8e2f458be7d95c ("Scan and Mesh Workshop"). After reading the schema and any returned instructions, tell me what this tool can do, we are going to be using it

The agent should fetch the tool schema first, collect the required parameters for your request, and then call the tool through AgentPMT.

Are my original files changed?

No. Every action that produces geometry returns new geometry_ids and leaves its inputs unchanged, so you can branch, compare and go back. Exports are written as new files in your File Manager.

Can I compare a 3D scan with the original CAD design?

Yes. compare_geometry can align the scan first, then reports mean, RMS, 95th percentile and maximum deviation, the share of the surface within your tolerance, inside or outside deviation for closed references, Chamfer, Hausdorff and F-score, and a colored deviation map: green within tolerance, red for extra material, blue for missing material.

Can I run my own Open3D code?

Yes. run_script runs Python with Open3D and NumPy preloaded on the geometries you pass in, inside an isolated worker with no network access. Call publish() to return geometry and report() to return values.

Can it repair a mesh for 3D printing?

Yes. process_mesh welds vertices, removes duplicate, degenerate and non-manifold elements, fixes triangle winding, repairs self-intersections and fills holes. Smooth hole filling follows the surrounding curvature and flags large patches as estimates. The quality report shows watertightness, holes and self-intersections before and after.

Can it turn a point cloud or depth camera frames into a mesh?

Yes. reconstruct_surface builds meshes with Poisson (closed and watertight), ball pivoting or alpha shapes, and rgbd_fusion fuses color and depth frames into a mesh using the camera intrinsics.

How are long jobs handled, and how long are files kept?

Results come back directly when the work finishes within about 20 seconds. Longer work returns a task_id: call get_task with wait_seconds 55 until it completes. Tasks survive restarts and wait in a queue when every engine is busy, and a job can run for up to 30 minutes. Generated files are kept for 7 days.

What can the AI segmentation recognize?

Pretrained Open3D-ML models label indoor scans (walls, floors, ceilings, doors, windows, tables, chairs and other furniture) and street-level scans (roads, sidewalks, buildings, vegetation, poles and vehicles), and find 3D boxes around cars, pedestrians and cyclists in a vehicle LiDAR sweep. Choose the scene and the matching model is picked for you, or name one of the 14 models directly.

Which files can I load?

PLY, OBJ, STL, OFF, GLB, glTF, PCD, XYZ, XYZN, XYZRGB and PTS meshes and point clouds, plus STEP and STP. Formats that do not store units (STL, OBJ, PLY, OFF, PCD and the XYZ family) need units when you load them. A job can take up to 2 GB of input files.

Scan and Mesh Workshop

Available ActionsEach successful request consumes credits as outlined below.

Details

Use Cases

Dynamic MCP Setup

30 Second Setup

Hosted Streamable HTTPS

Config Example

Frequently Asked Questions

How do I connect this tool to an external agent?

How does an external agent use this tool?

Are my original files changed?

Can I compare a 3D scan with the original CAD design?

Can I run my own Open3D code?

Can it repair a mesh for 3D printing?

Can it turn a point cloud or depth camera frames into a mesh?

How are long jobs handled, and how long are files kept?

What can the AI segmentation recognize?

Which files can I load?

Looking for help integrating AI into your business? Set up a free consultation.