ChatGPT Subscription MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CODEX_HOME | No | Controls Codex sign-in and state location. Default is ~/.codex. | |
| CHATGPT_MCP_MODEL | No | Optional model override. Default is the account default from model/list. | |
| CHATGPT_MCP_DATA_DIR | No | Data directory for completed responses, image files, copied reference inputs, job metadata, and conversation mappings. Default is ~/.chatgpt-subscription-mcp; use an absolute path. | |
| CHATGPT_MCP_CODEX_PATH | No | Optional absolute path to a native Codex executable. Defaults to the bundled Codex runtime. |
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 | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| chatgpt_statusA | Check subscription login, available text models, usage limits, and bridge setup without generating content. |
| chatgpt_loginA | Begin ChatGPT browser sign-in when the user asks to connect their account. Returns a URL for the user to open; never ask for a password or token in chat. |
| chatgpt_askA | Ask the subscription-backed OpenAI assistant a question or continue its conversation. Uses plan limits. Returns a background job; poll chatgpt_job for the answer. |
| chatgpt_generate_imageA | Generate a real image, or edit reference images, with built-in OpenAI image generation using ChatGPT subscription limits. Describe size, composition, transparency and edits in the prompt. Returns a background job; poll chatgpt_job. No image is promised until the job completes. |
| chatgpt_jobA | Retrieve a generation job, response text, conversation ID, and image paths. Completed small images are returned inline. Poll running jobs; do not resubmit the original request. |
| chatgpt_cancelA | Cancel a queued or running generation. Usage already consumed is not refunded. Poll the job to confirm terminal status. |
| chatgpt_read_imageA | Return a previously generated image inline using job and image IDs. Only reads images produced by this bridge, up to 4 MiB per image. |
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 7 tools
Each tool targets a distinct phase of the bridge workflow: status, login, cancel, ask, generate_image, job polling, and read_image. The only mild overlap is between chatgpt_job (returns job metadata, text, and image paths) and chatgpt_read_image (returns an image inline), but the descriptions clearly delineate their purposes.
All tools share the chatgpt_ prefix and snake_case formatting, which is highly predictable. The only minor deviation is that some names are action verbs (ask, generate_image, read_image) while others are nouns (status, job), but the convention remains readable and consistent overall.
Seven tools is well-scoped for a subscription bridge, covering authentication, generation, polling, cancellation, and image retrieval without bloat or obvious redundancy.
The surface covers the full core lifecycle: login, status checks, content generation (text and image), job polling, cancellation, and image retrieval. Minor gaps exist, such as no explicit logout or conversation-listing/management tool, but agents can work around these.