comfy-colab-kit
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CKIT_URL | No | Comfy Colab Kit gateway endpoint URL (e.g. https://comfy.example.com) | http://127.0.0.1:8080 |
| CKIT_TOKEN | No | Optional bearer token for authenticating with the comfy-gate instance |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
| prompts | {} |
| resources | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| gate_statusA | Inspect the Comfy Colab session state, active profile, progress, remaining time, and Compute Unit balance. |
| gate_wakeB | Wake or join the shared Google Colab L4 GPU instance with a specified model profile (e.g. 'moss-image', 'sd-latest', 'qwen-image'). |
| gate_txt2imgB | Generate an image from a text prompt using the active profile's pipeline (MOSS-Image DeBand or SD 3.5 Large fp8). Returns full image URLs. |
| gate_img2imgB | Generate a new image variation from a reference image (local file path, HTTP/HTTPS URL, or uploaded ComfyUI input filename) and a text prompt. |
| gate_generateB | Submit a raw ComfyUI prompt workflow graph to the running instance, wait for execution, and return image URLs and metadata. |
| gate_stopA | Stop the Colab GPU session immediately and release hardware to avoid unnecessary Compute Unit charges. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 6 tools
Each tool maps to a distinct action: status inspection, session start, three generation modes, and stop. The only mild overlap is gate_txt2img vs gate_generate, but the descriptions clarify that one uses a preset profile pipeline while the other accepts a raw ComfyUI graph.
All tools share a consistent 'gate_' prefix and snake_case form, which makes the set instantly recognizable as one family. The only deviation is gate_status being a noun while the rest are verbs (wake, txt2img, img2img, generate, stop).
Six tools is a tightly scoped set for a GPU session manager plus image generation surface. Every tool earns its place, with no redundant or filler entries.
The session lifecycle is fully covered (status, wake, stop) and generation is well served by txt2img, img2img, and raw-graph modes. Minor gaps exist: no tool to enumerate available model profiles (they are only referenced by example) and no way to cancel or interrupt an in-flight generation.