AI copilot for TouchDesigner that edits live node graphs
TDPilot by Dreamrec is an AI copilot for TouchDesigner that connects MCP-capable language models to a live project. It lets agents inspect, create, wire, and modify operators through natural-language commands, turning prompts into in-application edits. The tool exposes a tailored MCP toolset and integrates model feedback with project state to help artists, creative coders, and technical directors reduce manual wiring and explore agent-guided experimentation within their existing creative workflow.
What tasks can you actually use it for?
The tool acts as an MCP bridge between language models and a live TouchDesigner session, so it generates concrete in-environment edits rather than isolated code snippets. It can build complex operator networks, wire nodes, troubleshoot errors, and tweak parameters using natural-language prompts. Typical tasks include rapid prototyping, iterative demo setups, and automating repetitive wiring. It exposes 114 specialized MCP tools that map to operator-level control and node parameters.
How reliable are the AI-driven changes and rollbacks?
TDPilot runs changes as transactions with snapshots and a multi-step rollback mechanism, so edits can be reverted if they produce unwanted results. The tool also offers live validation and bounded repair routines that check network state before committing changes. Deterministic visual techniques and truthful concept routing reduce mismatched edits by aligning generated actions with current project state. Transactions and an activity log provide a safety net for iterative agent workflows.
What inputs and environment does it require?
TDPilot requires a running TouchDesigner instance and an MCP-compatible client such as Claude Desktop or Cursor; it typically deploys as a local server using Node.js or Python environments. The tool includes an OCR sidecar and a trace viewer to give visual feedback and assist debugging when agents inspect the UI. File handling follows what TouchDesigner accepts, because the model operates by reading and mutating the live node graph rather than processing static project files.
Is it practical to add to a creative workflow?
Adopting the tool requires familiarity with TouchDesigner's node logic and with running a local MCP client, so there is a setup and learning curve for non-technical artists. The extensive operator-level toolset enables fine-grained automation but demands that prompts precisely describe desired node arrangements. For privacy, the architecture processes interactions locally on the user's machine, which reduces external data transfer. The trace viewer and OCR sidecar shorten debug cycles by exposing visual state to assisting models.
A focused option for teams already using Pilot Studio tooling
TDPilot is a pragmatic option for teams already using other Pilot Studio tools, because Dreamrec's portfolio includes LivePilot and ComfyPilot that share design assumptions. This continuity reduces integration friction for organizations standardizing on the same agent-driven tooling. Adopt it when the intent is to incorporate agentic editing into existing pipelines; allow time for workflow adaptation and policy definition before production use.
