Skip to main content
Glama

add_fact

Record ONE thing you now know about the athlete, with how long it is true. This is the assistant's memory: it is NOT a copy of the snapshot and NOT a training log. Three tests, all must pass — (1) it is not already handed to you every turn (age, weight, hours, available days, thresholds, zones, race dates, the plan, load, readiness, response_profile), (2) no tool computes it (get_metric_history, get_records, get_activity), (3) it outlives this session — the verdict on one workout goes on that workout with set_activity_notes.

AREA decides what survives when context is tight: "health" and "constraint" are never cut, "pattern" is the only capped area. Nuance goes in tags, which are free.

CONSTRAINTS include a fixed commitment on a day the profile calls free, another coach's remit, an injury to plan around, and EQUIPMENT — tagged equipment, recorded as what they actually own in their words, with a horizon like anything else (a gym membership is an until, a turbo in the hallway is often a season).

HORIZON is required and there is no "unknown": if you cannot say how long it holds, ASK before writing. "stands" = until they say otherwise. "until" + dueDate = it ends (a race, an arrangement); when the date passes the fact stops applying and its opposite does NOT start applying. "recheck" + dueDate = it does not become false, it becomes untrustworthy (a threshold, a blood value, weekly hours). "season" + seasonFrom/seasonTo = true only part of the year (months, may wrap: 11-03 is November to March). "pattern" = an observation; set confirmations as you see it again.

dueAction is the line between a date that becomes a CALENDAR item and a date that is merely a question. Set it only when the due date carries something the athlete DOES ("сдать ферритин", "перемерить порог 30-мин тестом") — that is proposed as a calendar entry on the date itself. Leave it empty for "ask whether this still holds": no calendar clutter for a conversation.

Silent when the athlete just told you the fact. A CARD when it carries a dueAction (it will put something in their calendar), when it is a pattern (that is your inference about them, not their words), or when it replaces a fact already on their page.

ONE IDEA, ONE LINE. If a live fact of the same area already says something like this, the call is HELD and you get those facts back with their ids: answer with new:true (genuinely different — both stay, and a different day, date, trip or measurement IS different however alike the wording), replaces:<id> (this supersedes it — the old one is retired as this is written), or confirms:<id> (literally the same thing — that fact is bumped and your line is discarded). Every fact in your snapshot already ends with its #id, so a pattern you see again is confirm_fact on that id, not a second add_fact.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
newNotrue = write it beside the similar facts add_fact showed you, because it is a different thing
areaYes
tagsNo
textYesthe fact itself, in the athlete's language, one thing only
sourceYesathlete = they said it; assistant = you inferred it
dueDateNorequired for until / recheck
horizonYes
confirmsNoid of a live fact that already says this — it is bumped instead, nothing new is written (only after add_fact held your call and showed it)
dueLabelNowhat it is tied to, e.g. '10 км'
replacesNoid of a live fact this one supersedes — retired as this is written; a card
seasonToNo
dueActionNowhat the athlete DOES on that date; set it and it becomes a calendar item
seasonFromNo
confirmationsNopatterns only — how many times you have seen it

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observed

TDQS

A4.1/5.0
Behavior1/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The annotations declare destructiveHint=false, meaning the tool performs only additive updates, but the description explicitly says `replaces:<id>` retires an existing fact as this one is written. That is a non-additive state change to existing data and contradicts the annotation. The rich behavioral detail about held calls, cards, and horizons cannot overcome the contradiction.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is front-loaded with purpose and organized into sections for area, constraints, horizon, dueAction, and duplicate handling. It is very long, but the length is largely justified by the tool's 14-parameter complexity and the absence of an output schema. Some repetition and all-caps rule statements could be tightened.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's complexity, 14 parameters, and no output schema, the description covers the full lifecycle: when to write, what not to write, horizon semantics, area behavior, card versus silent behavior, and how held calls return similar fact ids. Nothing critical for correct invocation appears to be missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

With 14 parameters and only 64% schema description coverage, the description adds substantial meaning beyond the schema. It explains horizon values (stands, until, recheck, season, pattern), dueDate requirements, dueAction versus mere questioning, tags, area survival rules, and the new/replaces/confirms semantics. This compensates well for the schema's coverage gaps.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific verb and resource: 'Record ONE thing you now know about the athlete, with how long it is true.' It immediately distinguishes this memory tool from the snapshot and training log, and later names sibling tools such as set_activity_notes and confirm_fact. An agent can tell what this tool is for without opening the schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Usage is richly specified: three tests must all pass before writing, computed values should use get_metric_history/get_records/get_activity, and workout verdicts belong in set_activity_notes. The description also explains when to be silent versus when to produce a card, and how to respond when a call is held with similar facts. This is explicit when-to-use, when-not-to-use, and alternative routing.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources