MCP App Store
by KeenEthics
GoodMem icon

GoodMem MCP Server for Claude and ChatGPT

By PAIR Systems

Overview

GoodMem: AI memory infrastructure. Python SDK and MCP tools manage embedders, spaces, memories via natural language.

How to connect

  1. Claude: Open the GoodMem plugin listing for Claude and install it in Claude Code. Claude listing
  2. ChatGPT: Open the GoodMem listing in the ChatGPT app directory and choose Connect. ChatGPT listing GoodMem exposes 20 tools here.

Tools

fetch

ChatGPT
Fetch the first page of original UTF-8 text for one search result. If truncated, continue with goodmem_memories_content using this id and next_offset. A not_indexed field means that memory never finished processing and is invisible to search — report that rather than presenting it as ordinary content.

goodmem_console_setup

ChatGPT
Create a one-time, signed-in link to the user's GoodMem console. Use quick_start for initial setup: the user enters their model provider API key on the console page, which configures an embedder, any supported chat model or reranker, and an optional first space. Use embedder, reranker, or llm for focused model configuration; embedder automatically opens Quick Start when none exists. Give the returned link verbatim and explain that it works once and expires after 30 minutes. After Quick Start, call goodmem_spaces_list and reuse a suitable space before creating another; otherwise retry the action that needed the model. Creating a new setup link invalidates any previous setup link for this connection.

goodmem_embedders_list

ChatGPT
List the embedding models configured on this GoodMem Cloud instance (used when creating spaces). To add one for the user, use goodmem_console_setup.

goodmem_llms_list

ChatGPT
List the LLMs configured on this GoodMem Cloud instance (used for synthesized answers). To add one for the user, use goodmem_console_setup.

goodmem_memories_content

ChatGPT
Read a page of original UTF-8 text. While truncated is true, call again with next_offset to read the rest. Offsets and max_bytes count UTF-8 bytes; pages end on character boundaries. Binary originals require retrieved passages or page metadata.

goodmem_memories_create

ChatGPT
Save one memory (plain text) into a space when the user asks to save or remember it. An upload or a request to read, summarize, or answer questions about a document does not by itself authorize persistent storage. For an authorized save of an attached file, use goodmem_memories_upload to preserve the original PDF, Office document, or other file. Use this text tool when the user wants extracted text, notes, or selected excerpts saved as memories. Embedding is asynchronous and does not block saving: keep going, and never poll each memory in turn. Check in bulk instead — goodmem_memories_list returns processing_status for memories in a space, up to its limit; use all=true for a complete listing. Check once partway through a long ingestion and once at the end. Only speak up about failure: if any row is FAILED, call goodmem_memories_get on that memory for processing_error and tell the user plainly what is broken and that the fix is in their GoodMem console. Do not narrate successful processing. Before retrieving, confirm the rows you need are COMPLETED. Set operation_id to recover the same resource when retrying after an interrupted response.

goodmem_memories_delete

ChatGPT
Permanently delete one memory. Destructive — only call after the user has explicitly confirmed deleting this specific memory.

goodmem_memories_get

ChatGPT
Get one memory's record, including its processing status; when processing has FAILED, processing_error carries the reason (for example an embedder whose provider key is invalid).

goodmem_memories_list

ChatGPT
List the memories stored in a space. Returns up to limit records (default 500). count is the number returned in this response; truncated: true means more memories exist beyond it — tell the user the count is partial. Pass all: true to walk the entire space and get the exact total. For 'how many' questions, pass count_only: true — the exact total with no records, safe on any size. Each row carries processing_status; failed_count on the response means that many memories in this space failed to process and are not searchable.

goodmem_memories_pages

ChatGPT
List the pages of an ingested document memory.

goodmem_memories_retrieve

ChatGPT
Semantic search across memory spaces, reranked by the instance's reranker when one is configured (reranker_error reports a reranker that failed — results then arrive in embedder order). Scope with space_ids, narrow with a metadata filter expression, and set answer=true only when a synthesized answer is wanted (slower). failed_count on the response means that many memories in the searched spaces never finished processing, so the results are incomplete — say so rather than presenting them as the whole picture. Report warnings, skipped_spaces, and synthesis_error. Per-space fallback interleaves local rankings and omits a combined answer; synthesize from the returned passages.

goodmem_memories_upload

ChatGPT
Save an original file attachment into a GoodMem space when the user asks to save or remember that file. Use this for PDFs, Microsoft Office documents, images, and other files GoodMem can process. An attachment or a request to read, summarize, or answer questions does not itself authorize persistent storage. Pass the native file object; do not transcribe the file or put base64 in the text create tool. Choose a unique operation_id before the first call. After an interrupted response, retry with the same operation_id, file_id, space, and metadata; download_url may be refreshed if it expires. A recovered upload reuses the same memory. If a known file size exceeds the limit below, ask for a smaller file or a split original document. If size is unknown, attempt the upload; report a size-limit rejection without retrying the unchanged file or substituting a transcription. File processing and embedding are asynchronous: accepted does not mean searchable. Use goodmem_memories_get or goodmem_memories_list to check processing_status; only COMPLETED memories are searchable. Report FAILED processing plainly. Set extract_page_images=true when PDF page images should also be retained. Maximum file size: 50 MiB (52428800 bytes).

goodmem_ping

ChatGPT
Check that the connected GoodMem instance is reachable and healthy.

goodmem_rerankers_list

ChatGPT
List the reranker models configured on this GoodMem Cloud instance. Rerankers re-score search results for higher precision; to add one for the user, use goodmem_console_setup.

goodmem_spaces_create

ChatGPT
Create a new memory space. If the instance has exactly one embedder it is selected automatically; otherwise pass embedder_name from the error's options. Set operation_id to recover the same resource when retrying after an interrupted response. If no embedder exists, return the Quick Start link to the user. After setup, call goodmem_spaces_list and reuse a suitable space before creating another.

goodmem_spaces_get

ChatGPT
Get one memory space's record by id.

goodmem_spaces_list

ChatGPT
List the memory spaces visible to the connected GoodMem account.

goodmem_spaces_update

ChatGPT
Rename a memory space.

goodmem_users_me

ChatGPT
Show which GoodMem account this connection belongs to.

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/goodmem"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/goodmem)