Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
70-tool MCP server for TouchDesigner: build, wire and debug live networks from an AI agent.
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
The maker has not listed install details in the registry. See the maker’s README below.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 7, 2026
The registry name was verified with the GitHub account dylanroscover, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
The registry does not list a required key, though some servers ask you to sign in on first use.
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (6.2.67) reached the registry 35 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code.
In the same category: Spec Driven Development, Remote Desktop Commander, Signoff
Servers in the same category from other makers, sorted by stars.
Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Five gates that make an agent's 'done' provable: run them, read the verdict, audit the record.
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
create at the speed of thought.
embody.tools · Documentation · Manifesto · Changelog
Embot — he hops through your network while Envoy builds
Embody puts your ideas on screen as fast as you can describe them. Operators, connections, parameters, the works. Want to try a different direction? Spin up a new approach in seconds. Compare attempts side by side. Branch off the one that works. The tool keeps up with you, instead of the other way around.
[Envoy](https://dylanroscover.github.io/Embody/envoy/) — *forward velocity.* An embedded MCP server lets Claude Code, Codex, OpenCode, Gemini, Cursor, Windsurf, and GitHub Copilot (via VS Code) talk directly to your live TouchDesigner session. Create operators, wire them up, set parameters, write extensions, debug errors — by saying what you want. No copy-pasting code. No describing your network in chat. Idea → operators in seconds.
Embody — *lateral velocity.* Tag any operator and Embody externalizes it to files on disk that mirror your network hierarchy. Try a new direction, branch off a good one, restore the state from yesterday — all in seconds. Your externalized files are the source of truth, so every project opens already in flow.
Convoy — *outward velocity.* Convoy-enabled Embody nodes on a trusted LAN discover, inspect, and control each other — one AI session relaying builds, test runs, saves, screenshots, files, and restarts to every machine in the room. A small per-user background app keeps each computer reachable even while TouchDesigner is closed, so a closed project can be started remotely. Convoy guide
[TDXN](https://dylanroscover.github.io/Embody/tdxn/) — *the substrate that makes it all possible.* TouchDesigner networks exported as human-readable YAML. The format is what lets your AI agent understand what's on the screen, what lets you diff one attempt against another, and what lets a network reconstruct itself from text on the next project open. TDXN is what makes the rest of this possible.
| What | Why it matters | |
|---|---|---|
| 🤖 | Envoy MCP Server | 70 tools let your AI assistant build, wire, parameterize, and debug live networks. The first time you watch it happen, you stop typing operator names by hand for good. |
| 📄 | TDXN Network Format | Networks become text. Diff two versions, revisit any version, hand an LLM a complete picture of what's on screen — all from a single .tdxn file. |
| 📦 | Automatic Restoration | Externalized files are written on save, so any COMP can be recovered from disk. By default (Export-on-Save) the .toe stays authoritative on open; switch to Roundtrip mode to rebuild TDXN-strategy COMPs from .tdxn on every open. |
| 📤 | Portable Tox Export | Pull any COMP out as a self-contained .tox with external references stripped. Ship a piece of your project anywhere. |
| 🐍 | Project Python Environment | One .venv per project, built against TouchDesigner's own interpreter — Envoy runs from it, your packages import inside TD from it, and AI agents manage it hands-free. Python Environment |
| 🛰️ | Convoy LAN Relay | Convoy-enabled Embody nodes on a trusted LAN discover, inspect, and control each other through Envoy — relay test runs, saves, screenshots, and restarts to other machines from one AI session. Convoy guide |
Requirements: TouchDesigner 2025.33230 or later (Windows / macOS). No Python setup needed — Embody builds a per-project Python environment (.venv) matched to TouchDesigner's own interpreter, and your own packages can live in it too. No special folder structure either: Embody works in any project folder, and if you happen to use git, every change is also a clean diff for free.
Download the Embody .tox from /release and drag it into your TouchDesigner project. The [Setup Wizard](https://dylanroscover.github.io/Embody/embody/setup-wizard/) opens and walks you through the choices that matter — how much autonomy Embody gets, what to externalize, whether to enable the AI assistant (Envoy) and for which tool, permissions, whether to join a trusted-LAN Convoy, and where config files live. Nothing changes until the final click, and you can re-run it anytime via the Setup Wizard pulse on the Embody COMP.
Updating Embody: Embody updates itself — pulse Check for Update on the About page (or set Auto-Update to check at startup), and a verified release is downloaded, backed up against, and swapped in place. Your settings and tracked externalizations live on disk and survive the update untouched. See the auto-update guide. Manual alternative: delete the old Embody COMP and drag the new
.toxin its place — the new version picks up your on-disk state automatically, no re-scan, no files rewritten.
Provisioning a machine with no one at it (a render node, a media server)? The bootstrap installs Embody into a .toe offline with TouchDesigner's own tools, in one line.
lctrl twice to open the tagger (pick a strategy for a COMP, a file format for a DAT)ctrl + shift + u to update all externalizations, or ctrl + alt + u to update only the current COMP. Externalized files are written on save; on open, the .toe stays authoritative by default (Export-on-Save), while Roundtrip mode also reconstructs TDXN-strategy COMPs from diskTip: Externalization is opt-in — nothing is written to disk until you tag it. To capture your AI assistant's work automatically, set Auto-Externalize New Ops (Envoy parameter page) and everything it creates through Envoy is tagged and externalized as it's built.
For supported formats, folder configuration, duplicate handling, Manager UI, and more — see the Embody docs.
Embody includes Envoy, an embedded MCP server that gives AI coding assistants direct access to your live TouchDesigner session.
Envoyenable) does the same thing with your current settings127.0.0.1:9870 (configurable via Envoyport; if the port is taken by another instance, Envoy scans forward automatically).mcp.json (STDIO bridge, so tools are available even before TD is running) at your AI project root. By default that's the git repo root; the wizard's config-location step — or the Aiprojectroot parameter — can point it at the .toe folder or a custom path instead. Projects without a git repo still get config generated in the .toe folder.mcp.json automaticallyThe generated config runs Envoy's bridged STDIO transport (recommended — it can launch and restart TD for you). If you'd rather wire a client by hand, the direct HTTP transport works whenever TD is running:
{
"mcpServers": { "envoy": { "type": "http", "url": "http://127.0.0.1:9870/mcp" } } }
| Tool | What It Does |
|---|---|
create_op | Create any operator type in any network |
set_parameter | Set values, expressions, or bind modes on any parameter |
connect_ops | Wire operators together |
execute_python | Run arbitrary Python in TD's main thread |
export_network | Export networks to diffable .tdxn YAML |
create_extension | Scaffold a full extension (COMP + DAT + wiring) |
get_op_errors | Inspect errors on any operator and its children |
...and 56 more. See the full tools reference.
When Envoy starts, it always generates an AGENTS.md file in your project root with TD development patterns and project-specific guidance. It also writes a client-specific config for whichever assistant you select in the Aiclient parameter (CLAUDE.md + .claude/ for Claude Code, opencode.json + .claude/ for OpenCode, Cursor/Windsurf rules, Copilot instructions, GEMINI.md for Gemini; Codex and OpenCode read AGENTS.md directly). For OpenCode and local-model setups, see the Local Models & Open Clients guide.
TDXN (TouchDesigner eXternal Network) is the file format that makes the rest of Embody possible. It exports an entire operator network — operators, connections, parameters, layout, annotations, DAT content — as a single human-readable YAML file. Your AI agent can read it. You can read it. Any text tool can diff it. The network can rebuild itself from it.
This is the substrate. Every other capability — AI-driven building, version control, automatic restoration — builds on top of it.
ctrl + shift + ectrl + alt + eexport_network / import_network MCP toolsSee the full TDXN specification for format details, import process, and round-trip guarantees.
| Shortcut | Action |
|---|---|
lctrl + lctrl | Tag or manage the operator under the cursor |
ctrl + shift + u | Update all externalizations |
ctrl + alt + u | Update only the current COMP |
ctrl + shift + r | Refresh tracking state |
ctrl + shift + o | Open the Manager UI |
| ctrl + shift + c | Copy the select
Only the beginning is shown; read the rest in the repo