constant-watch
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CONSTANT_WATCH_DATA | No | Override the storage root for the Constant Watch journal. Defaults to ~/Library/Application Support/Constant Watch/ on macOS or %LOCALAPPDATA%\Constant Watch on Windows. All clients reading the same journal must use the same data root. | |
| CONSTANT_WATCH_HELPER | No | Override the helper executable for command-line operation. |
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": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| read_reviewA | Read an evidence-linked review and editable status draft for an inclusive YYYY-MM-DD range (up to 31 days). Counts are observations, not time worked. Projects are suggested groups; no completed actions are inferred. |
| read_day_flowA | Read one continuous daily Markdown journal across all apps, grouped into chronological sessions. day is YYYY-MM-DD. |
| day_sessionsC | Read chronological app/window sessions, with summaries and source evidence. Use next_offset for the next page. |
| list_appsA | List captured applications, bundle IDs, observation counts, and available date ranges. |
| search_screen_memoryB | Search captured text and summaries. All query words must match. Optionally filter by bundle ID and YYYY-MM-DD. |
| recent_activityB | Read latest observations with source text. Pass the last id as before_id to page backward. |
| read_app_dayB | Read a complete app-specific daily Markdown journal. day is YYYY-MM-DD. |
| ask_memoryB | Answer a natural-language question with exact captured quotes and source citations. Optional YYYY-MM-DD and app/topic filters. Say evidence is missing when no relevant record exists. |
| list_topicsC | List suggested project/topic groups across apps, with counts and grouping notice. |
| read_topicA | Read chronological evidence across apps for an exact topic key returned by list_topics; paginate using next_offset. |
| read_observationA | Read a cited observation's original Accessibility/OCR text and source URL. Does not open links or take actions. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| apps_resource | Application index for local screen memory. |
TDQS
Scored across 11 tools
Several tools overlap heavily around the same daily-scoped retrieval: read_day_flow, day_sessions, and read_app_day all surface chronological day data, and recent_activity is close to day_sessions. read_review also overlaps with read_day_flow, and search_screen_memory vs ask_memory share retrieval intent, though descriptions do give useful distinguishing cues.
Most tools follow a clear snake_case verb_noun pattern (list_apps, read_observation, read_review, search_screen_memory, ask_memory, list_topics). A few nouns-first names break the pattern (day_sessions, recent_activity), but naming remains readable and largely predictable.
11 tools is well-scoped for a capture-and-recall memory server. Each tool maps to a distinct retrieval entry point (apps, observations, days, sessions, search, Q&A, topics) without obvious padding.
As a read-only observation browser, the surface is fairly complete: discovery (list_apps, list_topics), retrieval (search_screen_memory, read_*), chronological views (day_sessions, recent_activity), and NL synthesis (ask_memory). Minor gaps like export or cross-day aggregation exist but core workflows are covered.