touchdesigner-agent

TouchDesigner Agent is an open-source TCP bridge created and maintained by Bo-Ning Li. It lets AI coding agents control TouchDesigner remotely — query nodes, write GLSL shaders, check compile errors, and iterate autonomously.

Project notes: boningli.com/blog/touchdesigner-agent

Why

TouchDesigner is powerful for real-time visuals, but its node-based workflow can be slow for shader-heavy projects. This bridge lets you describe effects in natural language, have an AI agent write the GLSL, and see results instantly in TD — with automatic compile-error feedback so the agent can self-debug.

Setup

1. TouchDesigner Side (one-time)

In your TD project:

1. Create a TCP/IP DAT — set Role = Server, Port = 7000
2. Create a Text DAT — name it callbacks, paste the contents of td-setup/callbacks.py
3. Link them — in the TCP/IP DAT's parameters, set Callbacks DAT = callbacks

The status should show listening on port 7000.

2. Agent / Python Side

from bridge.client import TDClient

td = TDClient()  # connects to localhost:7000

# Query any TD state
td.query("op('/project1/glsl1').type")          # → 'glsl'
td.list_nodes("/project1")                       # → ['glsl1', 'noise1', ...]

# Write a compute shader directly into a DAT
td.write_glsl("/project1/glsl1_compute", """
void main() {
    const uint id = TDIndex();
    // ...your GLSL here...
    P[id] = vec3(0.0);
}
""")

# Check if it compiles
result = td.glsl_check("/project1/glsl1")
print(result["ok"])      # True / False
print(result["errors"])  # compile error string if any

GLSL Workflow

All shader code lives as local .glsl files in a shaders/ folder inside your TD project directory — synced to TD automatically. This way everything stays in Git and can be edited by any tool (other AIs, text editors, etc.) without opening TD.

One-command setup

# Creates {project_dir}/shaders/particle_compute.glsl + a synced Text DAT in TD
td.setup_shader('particle_compute', project_dir='C:/Projects/my_td_project', initial_code="""
void main() {
    // your GLSL here
}
""")

This does three things:

1. Creates shaders/particle_compute.glsl in your project directory
2. Creates a Text DAT (particle_compute) in TD pointing to that file
3. Enables sync mode — TD auto-reloads whenever the file changes

Editing later

# Just write to the file — TD picks up changes automatically
td.write_glsl_file('C:/Projects/my_td_project/shaders/particle_compute.glsl', new_code)

# Check if it compiles
result = td.glsl_check('/project1/glsl1')
print(result["ok"], result["errors"])

Protocol

The TCP server accepts single-line text commands and returns single-line JSON:

Command	Description	Response
<expression>	Evaluate Python expression in TD	{"result": "..."}
exec:<code>	Execute Python statement	{"ok": true}
glsl_check:<path>	Refresh inputs + return compile errors	{"ok": true/false, "errors": "...", ...}

File Structure

bridge/
  client.py           # Python client for external agents
td-setup/
  callbacks.py        # Paste into TD's Text DAT
CLAUDE.md             # Agent instructions

License

MIT