Anamnesis MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| OPENAI_API_KEY | No | OpenAI API key for remote embeddings. If not provided, local Ollama embeddings are used. | |
| ANAMNESIS_STORE | No | Path to the Anamnesis data store directory. | ~/.anamnesis |
| EMBEDDING_MODEL | No | The embedding model to use. Options: text-embedding-3-small (remote) or nomic-embed-text (local Ollama) | text-embedding-3-small |
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
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| extensions | {
"io.modelcontextprotocol/ui": {}
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| anamnesis_pingA | Quick health check — returns server status and whether the embedding model is loaded. Use this to verify the MCP server is running before calling heavier tools like anamnesis_recall. |
| anamnesis_scan_sourcesA | Discover Claude session and plan files that have not yet been imported into Anamnesis. Returns a paginated list of files with their paths, types, projects, and line counts. Use this as the first step when importing the user's conversation history into memory. Call again with offset to get more. |
| anamnesis_split_sourceA | Register a source file in the Anamnesis database, split into chunks by line range. Call this after anamnesis_scan_sources to break a large file into manageable pieces for reading and memory creation. Each chunk becomes a source record that can be fetched and processed into a memory. |
| anamnesis_fetch_sourceA | Read the raw text of a source chunk by its ID. Returns the actual conversation or plan text from the file at the registered line range. After reading and understanding the content, call anamnesis_create_memory to store a summary as a memory. |
| anamnesis_create_memoryB | Store a new memory that you have crafted from reading a source or from the current conversation. The summary should be 2-3 focused sentences emphasizing keywords for future recollection. Include relevant technology names, pattern types, and problem descriptions. If source_id is provided, the source chunk will be marked as processed. |
| anamnesis_recallA | Search your long-term memory by semantic similarity. Use this at the start of any non-trivial problem, when the current context feels familiar, or when the user asks what you remember about a topic. Pass a natural-language description of what you're working on or looking for. |
| anamnesis_searchB | Keyword and tag search across memories. Faster than recall when you know specific terms to search for. Supports filtering by tags and project. |
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 has a distinct role in a clear import-then-recall pipeline (scan → split → fetch → create_memory, plus recall/search/ping). The only mild overlap is anamnesis_recall vs anamnesis_search, but the descriptions clearly differentiate semantic vs keyword/tag lookup.
All tools share the anamnesis_ prefix and use a verb_noun pattern (fetch_source, scan_sources, split_source, create_memory). Minor deviations are the single-word ping and recall/search, which are still readable and consistent in style.
Seven tools is well-scoped for a memory store: a coherent import pipeline plus retrieval and a health check, with each tool earning its place and no redundancy.
The ingest-and-retrieve lifecycle is largely covered, but memory lifecycle operations are missing—no update_memory, delete_memory, or fetch_memory-by-id—so agents cannot correct or prune stored memories. These are notable gaps, though the core workflow still functions.