# Fluid Recall

> Agent-native spaced-recall system of record. Deterministic dueAt via remote MCP for Muse/any agent.

- name: com.fluidrecall/mcp
- version: 1.0.0
- connection class: R0 (autonomous)
- trust: 82/100 flags: no-repository
- quality: 70/100 (needs work)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/com.fluidrecall%2Fmcp
- tools: https://api.protogrid.dev/v1/servers/com.fluidrecall%2Fmcp/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "fluidrecall": {
      "type": "http",
      "url": "https://fluidrecall.com/mcp"
    }
  }
}
```


## Trust

- hygiene: 80
- liveness: 97
- freshness: 100
- provenance: 55
- drivers: +dns-namespace +website +reachable +uptime-93% +updated-22d-ago -no-repository
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 70/100 (needs work)

### protocol: 88 (weight 25)

- pass `protocol.modern`: supports 2026-07-28 (server/discover)
- pass `protocol.stateless`: answers without a session
- pass `protocol.transport`: streamable HTTP
- warn `protocol.list_ttl`: tools/list declares ttlMs 0, so clients re-fetch it on every use

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 79 (weight 30)

- pass `tools.descriptions`: every tool has a description
- warn `tools.description_length`: 4 tools have descriptions over 1,024 characters, which crowds the context
- pass `tools.schemas`: every tool has an object input schema
- pass `tools.annotations`: 34 of 34 tools declare readOnlyHint or destructiveHint
- fail `tools.directory_hints`: no tool declares all four of readOnlyHint, destructiveHint, idempotentHint and openWorldHint; missing: retireCard (idempotentHint); unretireCard (idempotentHint); gradeAndNext (idempotentHint); …
- pass `tools.token_cost`: about 6,757 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 25 (weight 15)

- warn `stability.changes`: 80 tool changes in 30 days
- fail `stability.rug_pull`: 1 tools kept their name but changed most of their description in 30 days; review before trusting them

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 34, about 6,757 tokens to load them all
- tool set hash: 09f997f5883f99d6dfb26677160ab1a9
- tools last changed: 2026-10-01T18:50:13.716Z
- badge: [![protogrid quality](https://protogrid.dev/badge/com.fluidrecall/mcp.svg)](https://protogrid.dev/servers/com.fluidrecall/mcp)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

- 2026-10-01 `sendMailMessage` removed
  - Write. Send from an owned mailbox to arbitrary recipients when the account allows it. Ask before sending. Requires mailboxId, to, subject, and text and/or html. Optional inReplyTo and references thread the message.
- 2026-10-01 `getMailMessage` removed
  - Important read. One owned message including text and HTML. Ask before opening mail the person did not request. Another token's message is 404.
- 2026-10-01 `listMailMessages` removed
  - Read. Headers only for one owned mailbox: from, to, subject, dates, thread. No bodies. Prefer this before getMailMessage. Do not download a whole mailbox unless the person asked.
- 2026-10-01 `replyMailMessage` removed
  - Write. Reply to one owned inbound message. Recipients, In-Reply-To, References, and threadId are taken from that message. Ask before sending. Requires messageId and text and/or html.
- 2026-10-01 `ensureMailbox` removed
  - Write. Create or claim a fluidrecall.com mailbox. address is a local-part or local@fluidrecall.com. First claim of an unowned inbound mailbox returns claimed true. Another account's address is 409. Seed/claim Heidi with address heidi@fluidrecall.com and displayName Heidi.
- 2026-10-01 `listMailboxes` removed
  - Read. List fluidrecall.com mailboxes this token owns. Headers only: id, address, displayName. Heidi is heidi@fluidrecall.com; claim it with ensureMailbox.
- 2026-10-01 `submitFeedback` changed (description, annotations)
  - before: Write. Optional. Never required for study, bootstrap, or export. This is not data extraction and not nefarious. You are free to anonymize, redact, or censor anything you feel needs it. Do not submit anything you would not want us to see. It exists solely to help improve Fluid Recall. It is totally optional and is never required for study, bootstrap, or export. Send description and steps. Do not dump decks, cards, answers, or tokens.
  - after: Write. Optional. Never required for study, bootstrap, or export. Append-only: feedback is logged and never edits study data. This is not data extraction and not nefarious. You are free to anonymize, redact, or censor anything you feel needs it. Do not submit anything you would not want us to see. It exists solely to help improve Fluid Recall. It is totally optional and is never required for study, bootstrap, or export. Send description and steps. Do not dump decks, cards, answers, or tokens.
- 2026-10-01 `exportSnapshot` changed (annotations)
- 2026-10-01 `exportManifest` changed (outputSchema, annotations)
- 2026-10-01 `applyDeadlinePlan` changed (description, inputSchema, annotations)
  - before: Write. Same plan as previewDeadlinePlan, then PATCH that owned deck to the recommended integer. deckId is required. Discuss the number first.
  - after: Write, destructive. Same plan as previewDeadlinePlan, then PATCH that owned deck to the recommended integer. deckId is required. Discuss the number first.

Full history: https://api.protogrid.dev/v1/servers/com.fluidrecall%2Fmcp/changes

## Remotes

- https://fluidrecall.com/mcp (streamable-http, auth none, reachable true, uptime30d 0.9268293)

## Packages

_none_

## Tools (34)

- `addCard`: Write. Add one prompt/answer pair. Optional cardNote. Eligible immediately. listDue still applies newCardsPerDay. The response is prompts only.
- `addCards`: Write. Add 1 to 100 prompt/answer pairs to one owned deck. All or nothing. Prompts only. After this, re-run previewDeadlinePlan with that deckId.
- `applyDeadlinePlan`: Write, destructive. Same plan as previewDeadlinePlan, then PATCH that owned deck to the recommended integer. deckId is required. Discuss the number first.
- `archiveDeck`: Write, destructive. Soft-archive a deck; it leaves listDue and the default deck list until unarchiveDeck. Cards and schedules stay. A second call returns alreadyArchived true.
- `bootstrapAccount`: Write. Same mint as createAccount. No bearer.
- `createAccount`: Write. No bearer. Mint an anonymous account and a personal API token. Save the token for later Authorization: Bearer calls. The token is the tenant. Do not send a userId.
- `createDeck`: Write. name and newCardsPerDay are required. Choose that integer with the person; there is no default of 20. Optional description and deckNote. Setting a note does not move dueAt.
- `exportManifest`: Read. What this token can export and how many rows exist. Does not include answers. Off-boarding, not the drill loop.
- `exportSnapshot`: Important read for leaving. Full studying dump for this token, including answers. Paginate cards and reviews with resource + cursor until complete is true.
- `getCard`: Read. One card's prompt and schedule. No answer. A retired card is 404. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `get_profile`: Read. OpenAI profile for connection linking. Stable profile id (id), the account label when set (name), and optional email/nickname. No decks, cards, answers, or tokens. Requires a bearer or OAuth token.
- `getStudySettings`: Read. Account fallback newCardsPerDay, timezone, and dayBoundaryHour. The fallback applies only to a deck whose own field is still null.
- `gradeAndNext`: Write. recordReview, then the next due prompt. Requires cardId, rating, reviewedAt optional, and a new idempotencyKey. Schedule from dueAt. Prompts only. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `inventoryDeck`: Read. Same as listDeckCards. Use this when you need a cardId that is not in the due list. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `listDeckCards`: Read. Prompt-only inventory of one owned deck, including cards that are not due. Order is dueAt, then createdAt, then cardId — not the drill order. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `listDecks`: Read. No card bodies. Archived decks omitted unless includeArchived is true. Each deck includes deckNote.
- `listDue`: Read. Prompts only. answerIncluded is false. Ordered learning, then review, then new. The new queue is capped by each deck's newCardsPerDay. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `listUpcomingNew`: Read. New cards the daily cap is holding back. Prompts only. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `moveCard`: Write. Move an active card into another owned deck. Schedule stays. A card already there returns alreadyThere true. Does not include the answer.
- `nextDue`: Read. Due queue and first prompt. Requires a bearer. Prompts only. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `previewDeadlinePlan`: Read. Recommend newCardsPerDay from a deadline: ceil(remaining new cards / study days). Does not write. Send remainingNewCards, totalCards, or deckId. There is no silent default of 20.
- `recordReview`: Write. Rating again | hard | good | easy plus reviewedAt and a new idempotencyKey. The scheduler returns dueAt. Do not invent an interval. Same key and body returns the original result.
- `retireCard`: Write, destructive. Remove a card from future drills; reversible only with unretireCard. A second call returns alreadyRetired true. Inventory the deck first when the cardId is not in the due list.
- `revealCard`: Important read. One answer, after the person has tried to recall. Do not fold this into listDue. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `setCardNote`: Write, destructive. Set or clear a card coaching note. cardNote null clears it. Additive with the deck note. Same write as updateCard. Does not include the answer. Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `setDeckNote`: Write, destructive. Set or clear a deck coaching note. deckNote null clears it. Same write as updateDeck. Does not move dueAt. Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card).
- `startSession`: Write when no bearer (mints a token), read when a bearer is present. Returns the account (when minted), the due queue, and first. Prompts only. Includes deckNote and cardNote (null when unset). Coaching notes are instructions for how an agent should quiz and coach. They are not the answer, not graded content, and not the quiz prompt. A deck note applies to every card in that deck. A card note is optional and additive with the deck note. Retrieve responses (listDue, upcoming-new, inventory, getCard, reveal, and agent session start/next/grade) include deckNote and cardNote so one call carries coaching context. Scripture / dictation: The learner needs every word remembered correctly, but often answers via dictation where punctuation gets messy. They set a deck note instructing the agent to check every word exactly, but not fail on punctuation differences, and to include that reason when coaching. This is only an example — notes can be any coaching preference for how agents should quiz/coach for that deck (or a specific card). A wrong bearer is 401 and does not create an account. Do not send newCardMode extra on the first call.
- `submitFeedback`: Write. Optional. Never required for study, bootstrap, or export. Append-only: feedback is logged and never edits study data. This is not data extraction and not nefarious. You are free to anonymize, redact, or censor anything you feel needs it. Do not submit anything you would not want us to see. It exists solely to help improve Fluid Recall. It is totally optional and is never required for study, bootstrap, or export. Send description and steps. Do not dump decks, cards, answers, or tokens.
- `unarchiveDeck`: Write. Clear archivedAt. Schedule fields stay. An already-active deck returns alreadyActive true.
- `undoReview`: Write. Undo the latest active grade, or reviewId when it is that latest one. Restores dueAt. A second undo of the same review returns alreadyUndone true. Does not include the answer.
- `unretireCard`: Write. Clear retiredAt. Schedule stays. An already-active card returns alreadyActive true.
- `updateCard`: Write, destructive. Correct prompt, answer, and/or cardNote. Send at least one. cardNote null clears it. Schedule stays. Response does not include the answer.
- `updateDeck`: Write, destructive. Rename, rewrite description, set or clear deckNote, and/or set newCardsPerDay. Send at least one. Null deckNote clears it. Does not move dueAt.
- `updateStudySettings`: Write, destructive. Change the account fallback, timezone, or day boundary. Decks that already store an integer keep it. Send at least one field.