Griptape Agent Config: Griptape Cloud
Run gpt-4o on a Griptape Cloud key
- CONFIG
This is the config node where Griptape eats its own dog food. The ComfyUI pack is made by the team behind griptape.ai, and this node points your agent at their own hosted platform - Griptape Cloud - using a GT_CLOUD_API_KEY instead of a provider key. As the changelog puts it: you no longer need a separate Griptape Cloud API key and an OpenAI key to run a simple agent. One key, one hosted brain, no local model to babysit.
How it works
Under the hood it's a thin config: it builds a DriversConfig containing a single GriptapeCloudPromptDriver, which proxies the model on Griptape's servers. The model dropdown currently has exactly one option, gpt-4o - the changelog says more models are coming, but as of now "Griptape Cloud" effectively means "gpt-4o, hosted by Griptape." It's the simplest config in the pack precisely because there's only one driver and one model to think about.
The key handling is the standard pack pattern: api_key_env_var defaults to GT_CLOUD_API_KEY, resolved from ComfyUI Settings → Griptape or the OS environment. Get the key at https://cloud.griptape.ai/configuration/api-keys.
The inputs that matter
Honestly, there aren't many:
model- one choice,gpt-4o.api_key_env_var(defaultGT_CLOUD_API_KEY) - the key name.use_native_tools(default on) - whether gpt-4o does its own tool calling or delegates to Griptape's engine.temperature,seed,min_p,max_tokens,response_format- the usual prompt-driver knobs.
Output is a single CONFIG, same socket as every other config node.
Installing it
One-pack install again:
cd ComfyUI/custom_nodes
git clone https://github.com/griptape-ai/ComfyUI-Griptape
or via ComfyUI Manager → search "Griptape". Restart, drop your GT_CLOUD_API_KEY into Settings → Griptape, and you're running a hosted agent. The pack's griptape[all] dependency and its torch install are the same shared gotcha - README's troubleshooting section is your friend if install breaks ComfyUI.
Should you use it?
If you're already a Griptape Cloud customer, this is obviously your config. If you're not, weigh what it buys you: a hosted gpt-4o that needs no local VRAM and no separate OpenAI key. Against that, you're paying Griptape's hosting markup instead of OpenAI's direct price, you get one model choice, and the whole thing lives in a commercial cloud - no local-data angle at all. For most people the OpenAI Drivers or OpenAI Compatible Drivers config is the better default. This node is for the person who wants "Griptape, the product, end to end."
Common issues
[WARNING]: GT_CLOUD_API_KEY is not set- the one failure mode you'll actually hit. The key must live in Settings → Griptape or the environment under that exact name; the node blocks with an ExecutionBlocker if the model driver can't authenticate.- Expecting more models - the single
gpt-4ooption surprises people. It's not you misconfiguring; that's just where the node is today.
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt_model_commentopt | STRING | Prompt Driver | — |
| modelopt | COMBO | gpt-4o | 1 options: gpt-4o |
| max_attempts_on_failopt | INT | 21–100 | Maximum attempts on failure |
| temperatureopt | FLOAT | 0.100–1 | Temperature for sampling |
| seedopt | INT | 10342349342 | Seed for random number generation |
| use_native_toolsopt | BOOLEAN | true | Use native tools for the LLM. |
| max_tokensopt | INT | -1 | Maximum tokens to generate. If <=0, it will use the default based on the tokenizer. |
| min_popt | FLOAT | 0.100–1 | Minimum probability for sampling. Lower values will be more random. |
| response_formatopt | COMBO | default | Format of the response |
| api_key_env_varopt | STRING | GT_CLOUD_API_KEY | Enter the environment variable name, not the actual API key |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| CONFIG | CONFIG | — |