teardown-mcpSAFE
Buildup: an MCP server that lets AI assistants build Teardown-ready voxel models (.vox) for mods. Pre-alpha.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
Status: pre-alpha (0.6.0). The MCP server works for simple vehicles and props: a fresh Claude Code session built a drivable car with it. See docs/ROADMAP.md.
Buildup is an MCP server that lets an AI assistant (such as Claude Code) build complete, Teardown-ready voxel models: multi-part .vox files using Teardown's material palette, previews the AI can look at to correct its work, and a manifest (object names, sizes, anchor points) plus a known-good XML skeleton, so the AI can then write the mod's XML and Lua files coherently.
Goal: anyone can describe a vehicle and get a working Teardown mod, without knowing voxel modelling or modding.
Using it with Claude Code
Requires uv and a clone of this repository. Buildup runs as a local MCP server over stdio:
uv run --project /path/to/teardown-mcp buildup-mcp --workspace /path/to/your/folder/workspace
The easiest way to register it for Claude Code sessions opened in a folder of yours (outside this repository) is:
uv run python scripts/write_mcp_config.py /path/to/your/folder
which writes /path/to/your/folder/.mcp.json. Start Claude Code in that folder, approve the buildup server, and ask for a vehicle. Buildup writes projects to workspace/projects/ and exported mods to workspace/mods//; copy a mod folder into Documents\Teardown\mods\ to test it. Buildup never writes into the game's folders.
Tools: projects (create_project, list_projects, project_summary, undo), vehicle templates to customize (start_from_template: sedan, SUV, pickup, van, truck), colors and parts (define_color, add_part, remove_part), drawing (draw_profile for silhouettes, draw_box, draw_cylinder, draw_ellipsoid, draw_wedge, cut_edges), operations (mirror_part, hollow_part, move_part, `add_whe
92fad725c56aOBSERVED · 2026-10-11Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control.
claude mcp add buildup-mcp -- uvx buildup-mcp
{
"mcpServers": {
"buildup-mcp": {
"command": "uvx",
"args": [
"buildup-mcp"
]
}
}
}Exposed tools (10)
5 read · 4 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.
| Tool | Risk | Description |
|---|---|---|
add_part | write | Add an empty part to the body (one named object in the .vox file). |
draw_box | write | Add, paint or carve a rectangular block of cells start <= cell < end. |
hollow_part | read | Empty the inside of a part, keeping a shell (fewer voxels, interiors). |
inspect | read | Describe the model in text: sizes, materials, colors and connectivity. |
list_projects | read | List the projects of the workspace. |
move_part | write | Move a part (a wheel |
project_summary | read | Show a project |
remove_part | destructive | Delete a part and all its voxels (undo brings it back). |
teardown_reference | write | Teardown facts needed to write the mod |
undo | read | Revert the last edits of a project (any tool that changed it). Cannot be redone. |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | WARN |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (4 observation(s))
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (2)
remove_part
return base64.b64decode(images[0].data)
Gates applied: no_behavioural_pass.
92fad725c56afull audit observations/trust-audit/mcp-server/snx830__teardown-mcp.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-11 | 92fad725c56a | SAFE | B | 89 | first audit |
Questions
What is the teardown-mcp MCP server?
Buildup: an MCP server that lets AI assistants build Teardown-ready voxel models (.vox) for mods. Pre-alpha.
What tools does teardown-mcp expose?
10 in total: 5 read-only, 4 that write, and 1 that can delete or overwrite (remove_part). Every one is listed on this page with its risk.
Is teardown-mcp safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B. Separately from the audit: 1 of its tools can destroy data, so scope the token you give it to what you actually need.
What credentials does teardown-mcp need?
No credential environment variables were found in its source, so it appears to need none.
How does teardown-mcp run?
It speaks stdio, so it runs as a local process your client starts. It is published on PyPI as buildup-mcp.
How current is this page?
The grade is for one exact copy of the source (92fad725c56a), read on 2026-10-11. The repository is watched and re-audited when it changes.