Anchrs + MCP

Bring your learning into the conversation.

Create cards, review what is due, and understand your practice with access you approve.

Anchrs uses the Model Context Protocol (MCP) to connect your learning account to compatible assistants. Preview cards before saving them, then review with the same schedule used by Anchrs.

See connection steps
Try asking

“Prepare five flashcards about recursion in my Algorithms deck. Show me before saving.”

“Help me review my due flashcards, one at a time, and save the ratings I choose.”

“Show my Anchrs study progress for the last 30 days.”

“How did my typing speed and accuracy change over the last 30 days?”

Connect your account

One account. Your choice of assistant.

Early access connection

MCP server URL

https://anchrs.com/mcp

Custom connections require setup with your assistant. Public directory availability has not been confirmed.

Ask about connection setup
  1. Choose a compatible client. It needs remote MCP over Streamable HTTP and OAuth sign-in. Your client must be registered with Anchrs before you connect.
  2. Add the MCP server URL. Use the endpoint shown here. Anchrs or your workspace administrator configures the client settings; hosted client secrets stay with that administrator. Account and workspace policies may limit custom connections.
  3. Sign in to Anchrs and approve access. Read the requested permissions: reading flashcards or typing data, creating and editing cards, or saving reviews. An older read-only connection needs fresh consent for writes. Sign in on Anchrs; never put your password or an access token in a chat.
  4. Enable the connection and ask a question. Start with a prompt above. Disconnect at any time in Anchrs Connections.

Client setup guides

For ChatGPT, follow the official custom MCP connection guide. For Claude, follow the official custom connector guide. Your administrator uses the registered OAuth client option.

Other MCP clients need a registered client ID, a matching callback address, and PKCE sign-in. Connection support depends on the client; directory approval is separate.

15 account-scoped tools

Prepare. Save. Review.

Compatible MCP Apps hosts can show a card preview with an explicit Save button and a review view with reveal and rating controls. Other clients use the same tools through conversation. Host compatibility is verified separately.

List flashcard decks

list_flashcard_decks

List the connected user’s flashcard decks, card counts, and unfiled card count. Use next_after_id to request the next page.

Find flashcards

list_flashcards

Read the connected user’s flashcards. Optionally filter by deck, text in the question or answer, or cards due for review. Text fields may be truncated at 12000 characters. Use next_after_id for pagination.

Get a flashcard

get_flashcard

Read a connected user’s flashcard question, answer, mnemonic, and latest review schedule by ID. Text fields may be truncated at 12000 characters.

Read typing sessions

list_typing_sessions

Read the connected user’s completed typing chunks and performance metrics, newest first, over a bounded UTC calendar-day window including today. Each row is one completed chunk, not a full source file. Source code and keystroke logs are excluded. Use next_before_id for pagination.

Get typing statistics

get_typing_stats

Summarize the connected user’s completed typing chunks over a UTC calendar-day window including today, with daily and per-language metrics. Accuracy is weighted by keystrokes; average WPM is weighted by completed chunks.

Connected Anchrs account

get_account

Identify the account selected by this connection. Returns a stable profile identifier and display label; takes no account selector.

List Anchrs decks

list_decks

List owned decks, including empty decks, with card counts. Use next_after_id for another page.

Review due flashcards

get_due_cards

Get owned flashcards due now, using Anchrs’ canonical oldest-card-first queue. Reveal answers only when the learner is ready. Submit the returned review_version with a rating to save progress.

Study progress

get_study_progress

Count owned cards and due cards and verified review events in a bounded UTC window. No retention or mastery estimate is available. Legacy unclassified events are reported separately. Deck rows paginate.

Preview cards before saving

preview_flashcards

Prepare a bounded set of front/back cards for an owned deck without persisting them. Display the destination and proposed content. Saving requires create_flashcards and an explicit user decision.

Create an Anchrs deck

create_deck

Create an empty private deck with a user-chosen name. Reuse the same idempotency_key for retries; a changed payload requires a new key.

Save flashcards to Anchrs

create_flashcards

Save 1–20 explicitly requested front/back cards in an owned deck and initialize their canonical review schedules. Confirm the intended content and destination before saving. When using preview_flashcards, use its returned idempotency_key. Reuse the key on retries and when reopening that preview; equal requests replay the receipt, conflicting reuse fails.

Edit a flashcard

update_flashcard

Replace the front and back of an owned card after the user approves the change. Requires its current review_version from get_flashcard or get_due_cards, preventing stale edits. Preserves scheduling history; retries use the same idempotency_key.

Save a flashcard review

submit_review

Record the learner’s rating using Anchrs’ existing SM-2-style scheduler: 0–1 incorrect, 2 hard, 3 recalled, 4 good, 5 easy. Use the version returned with the card; stale versions fail. Never infer a rating without the learner’s answer. Reuse the same idempotency_key on retry. Internal scheduling fields are not accepted.

Finish a review session

complete_review_session

Acknowledge completion using only the review IDs successfully saved in this session. Validates owned review events and records content-free completion analytics once for this set. Does not change schedules.

Keep the measurement honest

Typing totals count completed chunks, not full files or unique practice runs. Accuracy is weighted by keystrokes; average WPM is averaged across chunks. Date windows use UTC.

Study progress counts verified review events recorded since this release. Older unclassified history is shown separately because it can include scheduling actions. Review counts are not a mastery score or retention forecast.

Clear permissions

Your learning data, with boundaries.

  • Only records you own are returned. Another user's deck or card ID does not grant access.
  • Flashcard access includes questions, answers, mnemonics, decks, review schedules, and study counts. Typing access includes completed-chunk metrics and aggregates. Account identification returns an opaque ID and an optional display name.
  • Typing source code, keystroke logs, typing review payloads, run tokens, and OAuth credentials are excluded from tool results. Text you put in a flashcard may be returned as flashcard content.
  • A preview does not save cards. With separate write permissions, your assistant can create or edit cards and save ratings you choose. Only a confirmed review submission updates Anchrs; an informal conversation does not save a review. No delete or typing-write tool is provided.
  • The assistant provider receives data returned by the tools you use. Its privacy and retention terms also apply. Disconnecting stops future access; it does not remove data already present in a conversation.

Read the Anchrs privacy policy or contact us with questions.

Building a client? The machine-readable connection guide includes public tool schemas, permissions, setup requirements, and configuration status. It contains no account data.