unified-memory-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| HERMES_HOME | No | Base for default paths (~/.hermes/*). | ~/.hermes |
| UM_DAG_FANIN | No | Nodes of a level collapse into one above. | 5 |
| UM_VEC_INDEX | No | auto (build in reindex, KNN when dim matches) | off (always brute force). | auto |
| UM_GRAPH_DECAY | No | mem_recall(hops>1): score multiplier per depth (depth-1). | 0.5 |
| UM_LINK_FANOUT | No | mem_recall(hops>1): max links per node per direction. | 20 |
| UM_ARCHIVE_PATH | No | Separate cold file. | ~/.hermes/unified_memory.archive.db |
| UM_ARCHIVE_BATCH | No | How many messages per archive pass. | 500 |
| UM_DATABASE_PATH | No | Path to the database. | ~/.hermes/unified_memory.db |
| UM_EMBEDDING_DIM | No | Skip probe dim (openai), useful offline. | |
| UM_CONTEXT_TOKENS | No | Effective host context window. | 200000 |
| UM_REDACT_ENABLED | No | Secrets gate on input (default ON — product is public). | true |
| UM_RETENTION_DAYS | No | 0 = keep forever. >0 = hot data older than N days moves to archive once a week (lossless; deletion only manual purge). | 0 |
| UM_SUMMARIZER_URL | No | LLM summarization endpoint; without it extractive is used. | |
| UM_ARCHIVE_SIZE_MB | No | Threshold for hot DB: oldest messages move to archive. | 1024 |
| UM_ASSEMBLY_BUDGET | No | Default token budget for mem_assemble. | 8000 |
| UM_EMBEDDING_MODEL | No | local: fastembed model strictly from registry; openai: passthrough name. | paraphrase-multilingual-mpnet-base-v2 |
| UM_RECALL_MAX_HOPS | No | mem_recall(hops): BFS traversal ceiling for links (above is error). | 3 |
| UM_REDACT_PATTERNS | No | Subset of catalog, comma-separated. | api_key,bearer_token,password_assignment,private_key |
| UM_EVIDENCE_PARTIAL | No | mem_evidence(cite): token coverage threshold for partial. | 0.5 |
| UM_FRESH_TAIL_COUNT | No | Fresh messages never compressed. | 20 |
| UM_SUMMARIZER_MODEL | No | Model for LLM summarization; without it extractive. | |
| UM_COMPACT_THRESHOLD | No | Fraction of window that triggers compaction. | 0.35 |
| UM_EMBEDDING_BACKEND | No | local (fastembed) | openai (8127/any OpenAI-compatible). | local |
| UM_EMBEDDING_TIMEOUT | No | HTTP timeout, seconds. | 30.0 |
| UM_EVIDENCE_MAX_REFS | No | mem_evidence: max refs per call (excess go to rejections). | 50 |
| UM_EMBEDDING_BASE_URL | No | Base for backend=openai. | http://127.0.0.1:8127 |
| UM_EVIDENCE_MAX_CHARS | No | mem_evidence: how many chars of body to take from each ref. | 8000 |
| UM_SUMMARIZER_API_KEY | No | Bearer token for the summarizer endpoint. |
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
Server capabilities have not been inspected yet.
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
No tools | |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
This server cannot be deployed
Maintenance
ActivityMaintained
ResponsivenessNo issues