anki-sync-mcp
A headless, sync-native Anki MCP sidecar that lets AI agents safely manage a real Anki collection without Anki Desktop or AnkiConnect, supporting sync with AnkiWeb or self-hosted servers.
README
anki-sync-mcp
A headless, sync-native Anki MCP sidecar for always-on AI agents.
anki-sync-mcp lets an MCP client safely manage a real Anki collection without Anki
Desktop or AnkiConnect. It runs as an authenticated Streamable HTTP service, uses Anki's
official collection and sync APIs, and synchronizes with AnkiWeb or a self-hosted Anki sync
server.
Unlike desktop bridges, it is designed to run unattended in Docker. Writes are serialized, idempotent, persisted across restarts, and synchronized before and after mutation. Destructive and schema-changing tools are disabled by default and use preview tokens plus verified backups when enabled.
It supports decks, notes, cards, tags, note types, templates, media, scheduling controls, backups, and explicit full-sync recovery.
Quick start with Docker Compose
Requirements: Docker Compose, Python 3 for token generation, and an AnkiWeb or self-hosted sync account.
git clone https://github.com/YOUR-USERNAME/anki-sync-mcp.git
cd anki-sync-mcp
mkdir -p secrets
python -c 'import secrets; print(secrets.token_urlsafe(32))' > secrets/anki_mcp_token
read -rsp "Anki sync password: " ANKI_PASSWORD
printf '%s' "$ANKI_PASSWORD" > secrets/anki_sync_password
unset ANKI_PASSWORD
chmod 600 secrets/*
export ANKI_SYNC_USERNAME="your-sync-username"
export ANKI_SYNC_HOST="" # empty for AnkiWeb; otherwise an HTTPS sync-server URL
docker compose up -d --build
docker compose exec anki-mcp anki-mcp-healthcheck
The included Compose configuration exposes MCP only to containers on its private assistant
network. Connect your agent container to that network and configure:
- URL:
http://anki-mcp:8000/mcp - Header:
Authorization: Bearer <contents of secrets/anki_mcp_token>
For a client running directly on the host, add this to the anki-mcp service in
compose.yaml, then run docker compose up -d again:
ports:
- "127.0.0.1:8000:8000"
Use http://127.0.0.1:8000/mcp as the MCP URL. After connecting, call
anki_sync_login once. The resulting sync host key is stored in the persistent Docker volume,
so you can later remove ANKI_SYNC_PASSWORD_FILE from compose.yaml and delete the password
secret when automatic reauthentication is not required.
Do not run two sidecars, Anki Desktop, or any other process against the same collection file. Keep destructive, schema, and full-sync feature flags disabled except during deliberate maintenance.
Configuration
Configuration is read from environment variables when the service starts. compose.yaml
supplies safe defaults and mounts the authentication token and initial sync password as Docker
secrets. Variables written as ${NAME:-default} can be overridden through the shell or a
Compose .env file. To change a fixed value, or use one of the settings not present in the
Compose file, add it under services.anki-mcp.environment.
Server and access
| Variable | Default | Description |
|---|---|---|
MCP_HOST |
0.0.0.0 |
Address on which the HTTP server listens. |
MCP_PORT |
8000 |
HTTP listen port (1-65535). Keep this consistent with expose, any ports mapping, and the client URL. |
MCP_PATH |
/mcp |
MCP endpoint path. It must be an absolute, static, non-root path and must not overlap /health. |
MCP_AUTH_TOKEN |
required alternative | Bearer token supplied directly. Use exactly one of this setting and MCP_AUTH_TOKEN_FILE; the Compose setup uses the file variant. |
MCP_AUTH_TOKEN_FILE |
required alternative | Path to a file containing the bearer token. One trailing newline is ignored. Compose mounts ./secrets/anki_mcp_token at /run/secrets/anki_mcp_token. |
MCP_SCOPES |
read,write,admin |
Comma-separated tool groups to register. Allowed values are read, write, admin, and destructive. Destructive tools require both the destructive scope and their feature flag. |
MCP_ALLOWED_HOSTS |
testserver,localhost,localhost:*,127.0.0.1,127.0.0.1:*,anki-mcp,anki-mcp:* |
Comma-separated HTTP Host patterns accepted by MCP DNS-rebinding protection. Add the service name or hostname used by remote clients. |
MCP_ALLOWED_ORIGINS |
http://localhost:*,http://127.0.0.1:* |
Comma-separated browser origins accepted by MCP transport security. Add explicit trusted origins when using a browser-based client. |
LOG_LEVEL |
INFO |
Uvicorn log level, such as debug, info, warning, or error. |
The /health/live and /health/ready endpoints are not bearer-authenticated. The supplied
Compose network keeps them private unless you publish the container port.
Collection and synchronization
| Variable | Default | Description |
|---|---|---|
ANKI_COLLECTION_PATH |
/data/collection.anki2 |
Local Anki collection file. Compose persists its parent directory in the anki_data volume; do not point multiple processes at the same file. |
ANKI_SYNC_USERNAME |
empty | AnkiWeb username or self-hosted sync username. Required for anki_sync_login and for bootstrap when no persisted authentication exists. |
ANKI_SYNC_PASSWORD |
optional alternative | Sync password supplied directly. It is optional after authentication has been persisted; do not set it together with ANKI_SYNC_PASSWORD_FILE. |
ANKI_SYNC_PASSWORD_FILE |
optional alternative | Path to the sync-password file. Compose mounts ./secrets/anki_sync_password at /run/secrets/anki_sync_password. Remove the setting and secret after login if automatic reauthentication is not needed. |
ANKI_SYNC_HOST |
empty (AnkiWeb) | Custom sync-server HTTP(S) base URL. Non-loopback servers require HTTPS; credentials, query strings, and fragments are rejected. |
ANKI_SYNC_ON_READ |
false |
Synchronize before coordinated reads. Individual read tools can still request a sync with sync_before=true. |
ANKI_SYNC_ON_WRITE |
true |
Synchronize before and after coordinated mutations. Disabling this permits local changes to remain unsynchronized. |
ANKI_SYNC_TIMEOUT_SECONDS |
300 |
Maximum wait for media synchronization (> 0, up to 3600 seconds). |
ANKI_BOOTSTRAP_MODE |
disabled |
Startup behavior: disabled, or download_if_empty to authenticate and perform a download-only initialization when the local collection is empty. Startup fails rather than uploading if the server requests a full upload. |
The persisted anki_data volume also contains synchronization authentication state, mutation
receipts, and backups. Preserve the volume across container recreation.
Safety and resource limits
| Variable | Default | Description |
|---|---|---|
ANKI_ALLOW_DESTRUCTIVE |
false |
Registers delete tools when MCP_SCOPES also includes destructive. Delete operations require a preview token and create a backup. |
ANKI_ALLOW_FULL_SYNC |
false |
Registers explicit full-download and full-upload recovery tools. It must also be enabled for schema-changing tools. |
ANKI_ALLOW_SCHEMA_CHANGES |
false |
Registers note-type and template mutation tools when ANKI_ALLOW_FULL_SYNC is also enabled. Note-type deletion additionally requires destructive access. |
ANKI_ALLOW_REVIEW_ANSWERS |
false |
Registers the administrative card-answering tool, which changes scheduling state. |
ANKI_CONFIRMATION_TTL_SECONDS |
300 |
Lifetime of destructive-operation preview tokens (30-3600 seconds). Tokens are single-use. |
ANKI_MAX_BATCH_SIZE |
50 |
Maximum notes accepted by one atomic batch-create operation (1-500). |
MCP_MAX_PAGE_SIZE |
100 |
Maximum limit accepted by paginated tools (1-1000). |
MCP_MAX_SEARCH_SCAN |
10000 |
Maximum collection items that bounded list/search and tag operations may scan (1-1000000). |
MCP_MAX_RENDERED_FIELD_BYTES |
262144 |
UTF-8 byte limit for each rendered text field in responses (1-4194304); oversized values are truncated and marked. |
MCP_MAX_CARD_FIELDS |
100 |
Maximum fields or templates returned for a note or note type (1-1000); omitted counts are reported. |
ANKI_MAX_MEDIA_BYTES |
1048576 |
Maximum decoded size of a media file read, written, or renamed (1-16777216 bytes). |
MCP_MAX_RESPONSE_BYTES |
1048576 |
Maximum serialized tool-result size (1024-16777216 bytes). It must be at least MCP_MAX_RENDERED_FIELD_BYTES + 4096. |
MCP_MAX_REQUEST_BYTES |
1048576 |
Maximum MCP request-body size (1024-16777216 bytes); larger requests receive HTTP 413. |
Boolean values accept the forms supported by Pydantic settings, including true/false,
1/0, and yes/no. Invalid or unknown configuration causes startup to fail instead of
silently falling back to a default.
Development
Requires Python 3.12 and uv.
uv sync --frozen --all-groups
uv run pytest
uv run ruff check .
uv run pyright
License
MIT — free to use, modify, and distribute under the license terms.
Trademark disclaimer
This is an independent, unofficial project. It is not affiliated with, endorsed by, or sponsored by Anki, AnkiWeb, Ankitects Pty Ltd, or their maintainers. Anki and AnkiWeb are trademarks of their respective owners.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.