memex
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| MEMEX_API_TOKEN | No | API token for Bearer auth (opt-in, off by default) | |
| MEMEX_AUDIT_LOG | No | Set to 'true' to enable audit logging (opt-in) | |
| MEMEX_AUTO_BACKUP | No | Set to 'true' to enable auto-backup (opt-in) | |
| MEMEX_AUTO_UPDATE | No | Set to 'true' to enable auto-update (opt-in) | |
| MEMEX_AUTO_CLEANUP | No | Set to 'true' to enable auto-cleanup (opt-in) | |
| MEMEX_ENCRYPTION_KEY | No | Encryption key for encrypting memory text at rest (opt-in) |
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 |
|---|---|
| memex_recallA | Search stored memories for |
| memex_storeA | Remember or store one new, self-contained memory verbatim -- exact text, not a summary. Optional valid_from/valid_to/event_at/source, and supersedes (ids this replaces; unknown or another user's ids are rejected). The reply echoes recorded_at and any declared window so you can confirm without a recall. Do not use this to correct something already in memory -- use memex_correct with the old id instead of storing a near-duplicate. |
| memex_correctA | Correct a memory that is wrong or outdated, in one call: stores |
| memex_getA | Fetch exactly one memory by id, verbatim, with its correction chain in both directions: what it replaced (oldest last) and what replaced it (newest last). Use this to resolve a "SUPERSEDED by " marker from memex_recall, or to confirm an id before passing it to memex_correct. An unknown id is an error. Do not use this to find a memory by topic -- use memex_recall instead. |
| memex_recentA | List memories by recency, no search query needed -- good for "what changed recently" or catching up after time away. |
| memex_healthA | Check whether the memory backend is up, and see store-wide stats (current/superseded/expired counts, oldest/newest record, queue and dead-letter depth) in one call. Call this after any other memex_* tool returns isError, before retrying. Do not call this to search memory -- use memex_recall for that. |
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 6 tools
Each tool has a clearly distinct role: recall searches, store adds, correct supersedes, get fetches by id, recent lists by time, and health checks backend status. There is no overlap, and the descriptions actively redirect misuse to the correct tool.
All tools share the memex_ prefix followed by a single lowercase word, forming a predictable memex_<operation> pattern. While recent and health are not verbs, the naming is uniform and immediately readable.
Six tools is a well-scoped size for a memory system: search, store, correct, retrieve, list, and health cover the core operations without redundancy. Each tool earns its place and the count feels neither thin nor heavy.
The surface covers the full memory lifecycle: storing, searching, retrieving by id, correcting with supersession, recency listing, and backend health. Deletion is intentionally omitted with supersession used instead, so there are no dead ends in the intended workflow.