Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/gluon/max-mcp-server/maxmspnpx skills add gluon/max-mcp-server --skill maxmspgit clone --depth 1 https://github.com/gluon/max-mcp-serverWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/gluon/max-mcp-server/maxmsp)<a href="https://agentmods.dev/skills/gluon/max-mcp-server/maxmsp"><img src="https://agentmods.dev/badge/skills/gluon/max-mcp-server/maxmsp.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00134 | $0.01432 |
| Opus 5 | $0.00067 | $0.00716 |
| Sonnet 5 | $0.00027 | $0.00286 |
| Haiku 4.5 | $0.00013 | $0.00143 |
Grade A, and why
maxmsp scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Max/MSP via maxmsp-mcp
You drive a running Max patch through the maxmsp-mcp tools. Objects you create
appear live on the canvas in max/mcp_host.maxpat, which the user keeps open.
The one rule
Call max_init before any other tool. Every other tool refuses until you
do. max_init returns the orientation guide — wire model, naming contract,
object cheat sheet, cookbook, and the layout and style rules. That guide is the
authoritative source on conventions; this file only tells you when to reach
for the MCP and how to stay out of trouble.
Prerequisites the user must satisfy
- The host patch is open.
max/mcp_host.maxpatmust be open in Max, and themax/folder on the Max search path (Options > File Preferences) so[v8 dump.js]resolves. If nothing appears in Max, this is almost always why. - Object reference (optional but recommended).
max_docneeds a local database built once from Max's own.maxref.xmlfiles:python tools/build_maxref_db.py /Applications/Max.app -o maxmsp_mcp/objects.json. Without it,max_docreports that no database is loaded; every other tool still works.
Working rules
- To build a patch from scratch, prefer
max_build_patch. Pass the whole graph — objects (each with anid, atype= the Max class such ascycle~, optionalargs, optional one-linecomment) and connections (byid). The server lays it out with no overlaps, wires it, verifies by reading back, and reissues any missing connection itself. You supply no coordinates. Use the unitary tools below only to EDIT a patch that already exists. - Address objects by their scripting name (
obj_0,obj_1, ...), returned by every create call. Wire, disconnect and delete by that name. Track the names you create; do not invent them. - When unsure of an object's signature — inlet/outlet order, argument types,
attribute names — call
max_doc <object>instead of guessing. In particular, before connecting to an inlet other than 0, confirm the object has that inlet ([*~]and[dac~]have only inlets 0 and 1). - Build incrementally, then verify: after wiring a non-trivial patch,
call
max_dump_patch(ormax_verify) to confirm every connection took, and reissue any that are missing. A failedconnectonly logs to the Max Console; it does not come back as an error, so reading back is how you catch it. - Before extending an existing patch, read it back first with
max_dump_patchso you build on what is really there, including objects the user added by hand. - Lay out per the guide, not by habit: build clear of the machinery (to the
right of, and below, the chat panel and
[thispatcher]), keep the signal flow vertical, and never overlap boxes. Follow the exact spacing in the guide returned bymax_init. - Do not propose
mxj-based solutions, and do not write malicious or unsafe-loud audio (see below).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 104 lines · 134 tokens per session scan A 51422e393fdf
maxmsp is a skill published in the GitHub repository gluon/max-mcp-server (19 stars, last pushed 2mo ago), licensed MIT. It adds 134 tokens to every session and 1,432 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
post-to-xhs
小红书内容发布技能。支持两种发布模式:(1) 上传图文模式 - 图片+短文;(2) 写长文模式 - 长篇文章+排版模板。 支持两种输入方式:用户提供完整内容和图片/图片URL,直接发布;或提供网页URL,自动提取内容和图片。 用户说"发长文"时使用长文模式,否则默认图文模式。.
sound-cues
Create and modify SoundCue assets — add/connect nodes (mixer, random, delay, attenuation, modulator, etc.) and set audio properties (SoundCueService). Use when the user asks to create a Sound Cue, build a SoundCue node graph, or wire audio playback logic.
AI Image & Video Generator — GPT Image 2, Seedance, ComfyUI
Generate images and videos from text with multi-provider routing — supports GPT Image 2.0 (near-perfect text rendering), Nanobanana 2, Seedream 5.0, Midjourney V8.1 (unified photorealistic + anime), Flux 2 Klein (cheap drafts), Seedance 2.0 / Veo 3.1 / Grok Video / Agnes Video, and local ComfyUI workflows. Includes…
image-prompting
Use when generating or editing images via blockrunimage — especially with GPT Image 2, Nano Banana, or Grok Imagine for posters, UI mockups, marketing assets, product shots, or anything with on-image text. Turns vague user requests ("make me a cool poster") into structured, text-accurate prompts that actually render…
Product Photoshoot Workflow
Multi-angle product imagery workflow. Use when the user wants to "shoot a product", "make e-commerce product images", "product photography set", "电商产品图", "产品多角度图", "brand product visuals", or provides a single product photo and asks for marketing-ready variations. Produces 4 distinct directions (lifestyle scene, macro…
video-perception
Use when the user mentions a video file (.mp4, .mov, .avi, .mkv, .webm), a YouTube URL, asks to watch/analyze/review a video, or references video content in conversation.