Skip to main content
Glama

AnkiWeb CLI (ankiweb)

Go Reference Documentation License: Apache-2.0 Author

A terminal-native Go CLI and MCP server for AnkiWeb, built by Sameer Bajaj. Canonical homepage & documentation: sameerbajaj.com/tools/ankiweb-cli.

Every existing Anki tool requires the desktop Anki GUI running with the AnkiConnect add-on bound to a local port. ankiweb talks directly to AnkiWeb's cloud services, enabling fast card creation with custom templates, batch Markdown importing, in-terminal flashcard reviews, deck triage, and community shared deck discovery — from any terminal or headless machine.


⚡ Highlights

  • Dynamic Custom Card Templates: Add cards to any deck respecting custom notetypes with arbitrary fields (--field "Additional Info=...", --field "Source=..."), not just basic Front and Back.

  • Markdown & Note Batch Importer (card import): Extract flashcards directly from Markdown reading notes, structured Q&A blocks, bullet pairs, and cloze syntax ({{c1::...}}).

  • Interactive Terminal Study (study): Review and grade due flashcards directly in the CLI with instant keyboard flips and 1–4 ease score recording (Again, Hard, Good, Easy).

  • Deck Backlog Triage (deck triage): Analyze review backlog debt ratios, detect bottleneck decks, and estimate clearance days across nested deck trees.

  • Cross-Deck Duplicate Detection (card duplicates): Identify duplicate notes and colliding prompts across multiple decks in your collection.

  • Community Deck Topic Briefing (shared brief): Surface top-rated community decks using Bayesian approval rate smoothing and audio/image media coverage filters.

  • Offline SQLite & FTS5: Fast local caching with full-text search across your collection.

  • AI Agent Native (MCP Server): Ships with a built-in Model Context Protocol server (ankiweb-mcp) and --agent JSON envelopes for autonomous assistants.


Related MCP server: Mousetail

📦 Installation

From Source (Go 1.22+)

# Install the ankiweb CLI
go install github.com/sameerbajaj/ankiweb-cli/cmd/ankiweb@latest

# Optionally install the MCP server
go install github.com/sameerbajaj/ankiweb-cli/cmd/ankiweb-mcp@latest

From Releases

Download precompiled binaries for macOS, Linux, and Windows from GitHub Releases.


🔐 Authentication

AnkiWeb uses session cookie authentication. You can authenticate in one of two ways:

  1. Environment Variable (Recommended for CI / Headless):

    export ANKIWEB_COOKIES="ankiweb=YOUR_SESSION_COOKIE"
  2. Interactive Login:

    ankiweb auth login
  3. Verify Auth Status:

    ankiweb auth status

Note: Public community shared deck searches (ankiweb shared search, ankiweb shared brief) require no authentication.


🚀 Quick Start

1. Check health & environment

ankiweb doctor

2. List your decks & due review counts

ankiweb deck list

3. Add a flashcard with custom template fields

# Basic card
ankiweb card add --deck "Default" --front "What is Raft?" --back "A distributed consensus algorithm" --tags "cs"

# Custom notetype with arbitrary fields
ankiweb card add \
  --deck "Computer Science" \
  --notetype "Technical Concept" \
  --front "Byzantine Fault Tolerance" \
  --back "Safety in presence of arbitrary failing nodes" \
  --field "Additional Info=Used in Tendermint and PBFT" \
  --field "Source=Castro & Liskov 1999" \
  --tags "cs,distributed-systems"

4. Batch import flashcards from Markdown

ankiweb card import --file study-notes.md --deck "Medicine"

Supported Markdown patterns:

  • Heading Q&A: ## Question followed by answer paragraph

  • Explicit Q/A: Q: Question / A: Answer

  • Bullet pairs: - Front :: Back

  • Cloze syntax: Text with {{c1::cloze deletion}}

5. Review flashcards in your terminal

ankiweb study --deck "Spanish" --limit 20

6. Triage deck review backlogs & pacing

ankiweb deck triage

7. Find duplicate prompts across decks

ankiweb card duplicates

8. Discover top-rated community shared decks

# Get a topic briefing with Bayesian approval ratings and audio coverage
ankiweb shared brief "Japanese"

# Search shared decks with media filters
ankiweb shared search "Spanish" --limit 20

📖 Command Reference

ankiweb card

  • add — Add a new note/card with support for custom fields (--field Name=Value)

  • get — Inspect note details and field values by note ID

  • search — Search your collection using Anki query syntax

  • notetypes — List available notetypes, fields, and decks in your collection

  • import — Batch parse and import cards from Markdown files or stdin

  • duplicates — Detect colliding prompts and duplicate notes across decks

ankiweb deck

  • list — List all decks with new, learn, and review queue counts

  • create — Create a new deck

  • rename — Rename an existing deck

  • delete — Delete a deck

  • limits — Inspect daily new and review card study limits

  • triage — Analyze backlog debt ratios and estimated clearance pacing

ankiweb study

  • study — Start an interactive or agent-automated review session

ankiweb shared

  • search — Search public community decks on AnkiWeb

  • info — View detailed ratings, samples, and reviews for a shared deck

  • brief — Generate a topic summary comparing highest-rated, freshest, and audio-rich decks

Utility

  • doctor — Run environment diagnostics and connectivity verification

  • sync — Sync collection data to local SQLite database for offline fast queries

  • sql — Run raw SQL queries against your local collection store

  • which — Natural language command discovery


🤖 Agent & Automation Usage

Every command supports non-interactive execution with structured outputs:

  • --agent: Enables non-interactive JSON envelopes with execution metadata.

  • --json: Formats output as JSON.

  • --select field1,field2: Projects only specific fields for minimal token usage.

  • --dry-run: Previews the network mutation payload without executing it.

# Dry-run preview of a card creation
ankiweb card add --deck "Default" --front "Hello" --back "World" --dry-run

# Structured agent query
ankiweb deck triage --agent

Model Context Protocol (MCP) Setup

Run the bundled MCP server in Claude Desktop, Cursor, or Codex:

{
  "mcpServers": {
    "ankiweb": {
      "command": "ankiweb-mcp",
      "env": {
        "ANKIWEB_COOKIES": "ankiweb=YOUR_COOKIE"
      }
    }
  }
}

👤 Author

Developed by Sameer Bajaj.


📄 License

Licensed under the Apache License, Version 2.0. See LICENSE for details.

Available Tools

37 tools
card_addC

Add or update a note/card in your AnkiWeb collection. Optional: deck, notetype, front (plus 2 more). Returns the new Card.

Inputs: back, deck, tags, front, +1 more.

card_duplicatesC
Read-only

Detect duplicate notes and colliding prompts across decks in your collection.

Inputs: csv, yes, deck, json, +19 more.

card_getC
Read-only

Get details and field values for a note. Optional: note_id. Returns the new Card.

Inputs: note_id.

card_importC
Destructive

Batch import flashcards from Markdown files, Q&A blocks, and cloze notes into any deck.

Inputs: csv, yes, deck, file, +22 more.

card_notetypesB
Read-only

List available note types and decks in your collection.

Inputs: no named parameters.

contextA
Read-only

Get API domain context: resource taxonomy, auth requirements, query tips, and unique capabilities. Call this first.

Inputs: no named parameters.

deck_createB

Create a new deck in your collection. Required: name. Returns the new Deck.

Inputs: name (required).

deck_deleteB

Delete a deck from your collection. Required: deck_id.

Inputs: deck_id (required).

deck_limitsB
Read-only

Get daily study and review limits for a deck. Required: deck_id.

Inputs: deck_id (required).

deck_listB
Read-only

List all decks with due, learn, and new card counts. Returns array of Deck.

Inputs: no named parameters.

deck_renameB

Rename an existing deck. Required: deck_id, name. Returns the new Deck.

Inputs: name (required), deck_id (required).

deck_triageC
Read-only

Analyze review backlog debt ratios, bottleneck decks, and estimated clearance days.

Inputs: csv, yes, json, agent, +18 more.

exportC
Destructive

Export data to JSONL or JSON for backup, migration, or analysis

Inputs: resource (required), id, csv, yes, +22 more.

importB
Destructive

Import data from JSONL file via API create/upsert calls

Inputs: input (required), resource (required), csv, yes, +21 more.

learnings_candidatesB
Read-only

List auto-captured improvement candidates awaiting judgment

Inputs: csv, yes, json, agent, +21 more.

learnings_confirmB

Confirm an open candidate: print its full payload, then materialize it

Inputs: id (required), csv, yes, json, +19 more.

learnings_forgetA
Destructive

Delete learnings matching a query (use --all to wipe every rule for that query)

Inputs: query (required), all, csv, yes, +22 more.

learnings_listB
Read-only

List recorded learnings from the local ankiweb search_learnings table

Inputs: csv, yes, json, agent, +25 more.

learnings_rejectB
Destructive

Reject a candidate and tombstone its derivation signature

Inputs: id (required), csv, yes, json, +19 more.

learnings_statsB
Read-only

Report learn-loop effectiveness metrics from the local events table

Inputs: csv, yes, json, agent, +18 more.

playbook_amendB

Append a note to an existing playbook (LLM-fired self-correction, silent)

Inputs: csv, yes, json, agent, +20 more.

playbook_listC
Read-only

List stored playbooks (query_family, content presence, last observed)

Inputs: csv, yes, json, agent, +18 more.

recallB
Read-only

Check prior learnings for a query before running discovery (LLM-fired, pre-discovery)

Inputs: query (required), csv, yes, json, +22 more.

shared_briefB
Read-only

Get a multi-metric digest for a topic comparing top-rated, audio-rich, and fresh community decks.

Inputs: csv, yes, json, agent, +20 more.

shared_infoB
Read-only

Get detailed information for a community shared deck. Optional: deck_id. Returns the SharedDeck. (public)

Inputs: deck_id.

sqlA
Read-only

Run read-only SQL against local database. Use for ad-hoc analysis, aggregations, and joins across synced resources. Requires sync first.

Inputs: query (required).

studyC
Destructive

Review and grade due flashcards interactively in your terminal with keyboard shortcuts.

Inputs: csv, yes, deck, json, +20 more.

syncC
Destructive

Sync API data to local SQLite for offline search and analysis

Inputs: csv, yes, full, json, +29 more.

teachB

Record a query -> resource mapping for future recall (LLM-fired, silent)

Inputs: csv, yes, json, agent, +28 more.

teach_lookupC

Install a manual entity-lookup row (kind, canonical, value)

Inputs: csv, yes, json, kind, +21 more.

teach_patternC

Install a manual generalization pattern (query_template, resource_template, entity_kind)

Inputs: csv, yes, json, agent, +24 more.

teach_playbookC

Record a CLI playbook + free-text notes for a query family

Inputs: csv, yes, json, agent, +23 more.

workflow_archiveC
Destructive

Sync all resources to local store for offline access and search

Inputs: csv, yes, full, json, +20 more.

workflow_statusB
Read-only

Show local archive status and sync state for all resources

Inputs: csv, yes, json, agent, +18 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    C2.9/5.0

    Scored across 37 tools

    Disambiguation3/5

    Several clusters overlap confusingly: sync vs workflow_archive both mean 'pull resources into local store', while search vs card_search vs sql all do querying with murky boundaries. The teach/teach_lookup/teach_pattern/teach_playbook/recall/learnings_*/playbook_* family is a dense maze of meta-learning tools an agent will struggle to choose between. Core deck_* and card_* tools are distinguishable, but the descriptions alone don't fully resolve the overlaps.

    Naming Consistency4/5

    The dominant pattern is resource_action (deck_list, card_search, shared_info, learnings_confirm, workflow_status), which is applied consistently across most families. A handful of bare verbs (sync, search, sql, study, teach, import, export) break the pattern, but they read clearly and map to obvious single-purpose actions.

    Tool Count2/5

    At 37 tools the surface is oversized for an Anki client, and a large fraction is internal agent-harness machinery (teach_*, learnings_*, playbook_*, recall) rather than Anki domain operations. A user/agent scanning this set must sift through many niche meta-tools to reach the core deck/card operations.

    Completeness3/5

    Deck lifecycle is well covered (create, list, limits, rename, delete, triage) and cards support add/update, get, search, import, and duplicate detection. However, there is no explicit card/note delete, no media handling, no programmatic review/scheduling control beyond the interactive study tool, and shared decks can be searched/viewed but not downloaded or imported — leaving real dead ends.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers