| id | memory-manager |
| name | Memory Manager |
| description | Store, search, and prune persistent agent memory. Supports semantic search, namespace isolation, and selective pruning to keep memory lean and relevant across sessions. |
| tags | ["memory","persistence","search"] |
| apiKeyRequired | false |
| commands | [{"name":"search","description":"Semantic search over stored memories","args":["query","limit"]},{"name":"store","description":"Persist a new memory entry","args":["content","namespace"]},{"name":"prune","description":"Remove outdated or low-relevance memories","args":["namespace","before_date"]}] |
Memory Manager
Persistent cross-session memory for an agent. Entries survive restarts and session clears.
Available Tools
| Tool | Purpose |
|---|
memory_store | Save a new memory entry with optional namespace tag |
memory_search | Full-text + semantic search across stored memories |
memory_list | List memories in a namespace |
memory_delete | Remove a specific memory by id |
Usage Patterns
Saving a Key Fact
Use memory_store to save:
content: "User prefers reports in Markdown with H2 section headers."
namespace: "preferences"
Recalling Context Before a Task
Before starting a long task, call memory_search with the task topic
to surface relevant prior context.
Post-task Cleanup
After completing a project, use memory_delete or memory_list + memory_delete
to remove stale entries that are no longer relevant.
Namespace Conventions
preferences — user preferences and style rulesfacts — domain facts learned during interactionstasks — task state that should survive session boundaries<project-id> — project-scoped memories
Notes
- Memory is stored in SQLite under
~/.agentflyer/ — no external service required.
memory_search uses BM25 full-text search; no vector embedding is needed.- Memory entries are private per gateway instance.
