LokalBot
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
LokalBot is the user's local memory of their meetings, screen, and what they shared about themselves. At the start of a session, call get_briefing. Before asking the user about past work, decisions, commitments, or where something stands, search LokalBot first.
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 |
|---|---|
| get_briefingA | Call this first in every session. Returns what LokalBot knows that this session may use: which sources are available, what the user shared about themselves, their open action items, current projects and goals, recent coding-agent work, and a timeline of recent days, weeks, and months. Check it, and search LokalBot, before asking the user about past work or decisions. |
| list_meetingsA | List recorded meetings, newest first. Returns id, title, date, and duration for each; include "summary" adds each meeting's summary, so one call covers several meetings. |
| get_meetingA | Fetch one meeting as markdown. Sections: metadata, summary, transcript. A long transcript comes in windows of max_characters; a cut window ends with the transcript_from value that continues it. |
| search_meetingsA | Word search across meeting titles, summaries, and transcripts. Meetings with every query word (any order, case- and accent-insensitive) come first, then partial matches; rare words weigh more, exact-phrase hits lead, and ties keep recency. Words of up to three letters or digits match only whole words. Wrap the query in double quotes for an exact phrase. Hits carry kind, snippet, and timestamp; one meeting gives at most five transcript hits. |
| ask_libraryA | Ask a question and get an answer synthesized by LokalBot's local model from the meeting library, with meeting citations — raw transcripts never leave this computer. Needs the LokalBot app running; the first call may take up to a minute while the model loads. |
| get_action_itemsA | List action items extracted from recent meetings, with the user's saved corrections and status applied. Overdue and soon-due work comes first. Due phrases are resolved to a date only when unambiguous (due_date); otherwise only the spoken phrase is returned. |
| list_peopleA | List people the user meets, joined from calendar attendee names, speaker names applied in LokalBot, and action owners. Returns names, meeting counts, last meeting, and open-action counts in each direction. Email addresses are never returned. |
| get_personA | One person's open commitments in both directions (what the user owes them and what they owe the user), recent decisions made in small meetings together, and shared meetings. |
| search_screenA | Search locally captured screen text and window titles. Returns text snippets and context metadata only; never pixels or encrypted file paths. Repeated captures of the same window and text count once. Requires the separate screen-memory permission. |
| get_timelineA | Get activity blocks and screen-context metadata for one local calendar day, oldest first. No pixels or encrypted file paths are returned. A list cut at the limit sets activity_truncated or screenshots_truncated; call again with after set to its last ended_at or captured_at to continue. Requires the separate screen-memory permission. |
| get_recent_activityA | Get recent app/window activity blocks, newest first. Requires the separate screen-memory permission. |
| get_app_usageA | Aggregate tracked time by app for one local calendar day. Requires the separate screen-memory permission. |
| get_screenshot_detailA | Get captured text and metadata for one context-moment id. It reports whether encrypted pixels exist but never returns decrypted pixels or a file path. Requires the separate screen-memory permission. |
| get_personal_contextA | Get what the user shared about themselves (facts, preferences, goals, principles) that is relevant to a task, most relevant first. Only items the user marked as shared are returned. Requires the separate personal-context permission. |
| search_filesA | Search the folders the user chose under Files (documents, PDFs, pictures, screenshots, recordings and video on their computer) by words. Returns matching files, best first, with the passage that matched. Requires the separate file-access permission. |
| search_agent_notesA | Search the notes the user's coding agents (Claude Code and Codex) saved for their own later sessions: project decisions, how-tos, preferences and pointers. Returns matching notes, best first, with their text. Use it to find what another agent or an earlier session already learned. Requires the separate agent-notes permission. |
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 16 tools
Most tools target a clearly distinct resource: meetings (list/get/search), people (list/get), files, agent notes, and screen memory each have separate entry points. The main soft spots are within the screen-memory family (get_timeline vs get_recent_activity vs get_app_usage vs search_screen) and the two meeting-answer paths (search_meetings vs ask_library), which an agent could plausibly confuse. Descriptions do enough work to resolve these in practice.
Every name follows a verb_noun pattern (get_, list_, search_, ask_) in snake_case throughout. The verb choices map predictably to intent: list_ for enumeration, get_ for single-resource fetch, search_ for query-based retrieval. No camelCase or stylistic drift.
16 tools is on the heavy side of the ideal band, but the server spans genuinely separate domains (meetings, people, action items, screen memory, files, agent notes, personal context), so the count is largely earned. Each screen-memory tool does something distinct rather than being a redundant variant.
Meetings have a full read lifecycle (list, get with pagination, search, synthesized Q&A) and people/action items/timeline have sensible entry points. Gaps are minor: search_files returns only a matching passage with no way to fetch a full document, and action items are list-only with no per-item fetch or status mutation from the agent side. These are workable around for a read-only memory server.