chatgpt
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CODEX_PATH | No | Override the path to codex.exe. By default, the executable path is found automatically after app updates. | |
| CHATGPT_MODEL | No | Override the ChatGPT model used by the server. | gpt-6.2-sol |
| CHATGPT_MCP_DATA | No | Override the location of the conversation index and empty Codex workspace. By default they live in data/. | |
| CHATGPT_TIMEOUT_SECONDS | No | Override the timeout in seconds. | 600 |
| CHATGPT_REASONING_EFFORT | No | Override the reasoning effort used by the server. | medium |
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 | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| chatgptA | Ask ChatGPT a question (runs in the background through the Codex CLI signed in with the user's ChatGPT plan; no API cost, does not touch the screen). Operations: 'ask' sends a prompt and returns ChatGPT's reply plus a conversation_id you can pass back to continue the same conversation; 'get_conversations' lists conversations started through this tool. ChatGPT here has no access to files, email, or other apps; include any text it needs in the prompt. |
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 1 tool
With only one tool there is no cross-tool ambiguity, so selection is trivial. The only friction is that 'ask' and 'get_conversations' are two distinct operations multiplexed behind a single tool name via an operations parameter, which the description does explain.
A single tool cannot be internally inconsistent, so there is no convention clash. However, the name 'chatgpt' is a product/service noun rather than a verb_noun action name like 'ask_chatgpt' or 'list_conversations', so it does not establish a predictable pattern.
One tool is borderline thin; splitting 'ask' and 'get_conversations' into two tools would be more discoverable. The narrow purpose (relaying a prompt to ChatGPT) does justify a minimal surface, so it is not a severe mismatch.
The stated purpose — ask ChatGPT and continue a conversation — is covered via prompt/response plus conversation_id, and listing conversations exists. Minor gaps remain (no delete/rename conversation, no model or parameter selection), but these are workaroundable or out of scope.