Reference/CLI

CLI

any is one binary in two roles: any run starts the server, and every other subcommand is an HTTP call to it. The CLI never opens the SDK or touches storage, so anything you can do from the terminal you can do from curl — and vice versa.

Global flags and exit codes

--addr <host:port>     # bind address for `run`; connect address otherwise — default: the address
                       # the running server recorded (<account-dir>/server.addr), else 127.0.0.1:7001
--control-token <hex>  # managed server's control token; prefer ANY_CONTROL_TOKEN (a flag is visible in ps)
--timeout <duration>   # request timeout, default 30s (does not apply to streams)
--verbose              # log the HTTP request/response to stderr
Exit code Meaning
0 success
1 user error — bad arguments, or a 4xx from the server
2 server error — a 5xx
3 transport error — the server cannot be reached

Output is pretty-printed JSON, always. The two deliberate exceptions are any file download (raw bytes to stdout) and streaming commands, which print one JSON object per SSE frame — {"event": "<name>", "data": <payload>} — so they pipe cleanly into jq. When the server is not running the CLI exits 3 and prints start it with any run in another terminal; there is no auto-start.

Input conventions: JSON-valued flags (--pipeline, --records, --draft, --body, --x-format, …) accept '<json>', @FILE or - for stdin; --filter is a Mongo-style filter object; windowed queries take --projection 'any,<typeId>' or '-_ver'.

Meta and auth

any init [--mnemonic "w1 … w12"] [--mnemonic-stdin] [--index N] [--new]
         [--config PATH] [--data-dir PATH] [--account ID] [--wallet PATH] [--passkey-stdin]
any run  [--config PATH] [--data-dir PATH] [--mode standalone|managed] [--account ID]
         [--addr host:port] [--wallet PATH] [--passkey-stdin]
         [--log-level debug|info|warn|error]
any auth login [--mnemonic …|--mnemonic-stdin|--account ID] [--index N] [--replace]   # POST /v1/auth
any auth logout                                                   # DELETE /v1/auth (managed)
any auth status                                                   # GET /v1/auth
any status                                                        # GET /v1/health
any stop [--config PATH] [--data-dir PATH] [--account ID]         # signal the server holding the account lock
any version                                                       # binary version, plus the server's when one runs

--mode managed starts a host-owned server: it never resolves an account from disk, the phrase arrives over POST /v1/auth on every launch, and it prints its control token as the second stdout line (CONTROL_TOKEN <hex>). auth login --replace (switch in place) and auth logout are managed-only and need that token (ANY_CONTROL_TOKEN). any stop sends no HTTP and takes no --addr: it finds the server by its held account lock and signals it — so it works against a wedged server and one on an ephemeral port; a standalone server refuses POST /v1/shutdown. Not available on Windows (no signal to send — stop the server with Ctrl-C).

any init creates the data dir and an account wallet, printing the BIP-39 mnemonic to stderr once. With --mnemonic / --mnemonic-stdin it restores an existing account (same phrase, same account id, a fresh device key) — prefer stdin so the phrase stays out of shell history. --index defaults to 1 (the any derivation index; 0 restores an anytype-derived account); --new forces an additional account. any run never creates wallets: with no resolvable account it starts unauthorized and waits for any auth login. See Accounts.

Account, identities, devices

any account set-metadata [--name N] [--description D] [--icon-cid CID]   # PUT /v1/account/metadata
any account redeem <code>                                               # POST /v1/account/access-code

any identities list | get <identity> | subscribe     # alias: any contacts

any devices list                                     # rows + active map + self
any devices register [--name N] [--app slug[=ver]]... [--remove-app slug]...
any devices activate <app>
any devices remove <peerId> --yes                    # permanent for that peer id
any devices query     [--filter J] [--sort K] [--limit N] [--offset N] [--total] [--projection P]
any devices subscribe [same flags]

GET /v1/account has no subcommand — call it with curl.

Spaces

any space get      <spaceId>
any space update   <spaceId> [--name N] [--description D] [--icon-cid CID]   # unset flag = keep, --name='' = clear
any space settings <spaceId> [--set k=v]... [--set-bool k=true|false]... [--set-num k=N]... [--unset k]...
any space delete   <spaceId> --yes
any space sync     <spaceId>                         # force one head-sync round
any space derived                                    # list well-known derived spaces
any space derived create <name>                      # materialize one (idempotent)
any space query      [--dataset spaces|profile] [--filter J] [--sort K] [--limit N] [--offset N] [--total] [--projection P]
any space subscribe  [same flags]                    # live space list, one frame per line
any datasets [<spaceId>]                             # dataset schemas (JSON Schema + x-scope)
any search <spaceId> <query> [--scopes basic,chat,props] [--limit N]
           [--mode hybrid|fts|vector] [--require T]... [--exclude T]...
           [--max-data N] [--passages N] [--filter J]

any space settings writes the account-private per-space settings object; push reads notifyMode from it: any space settings $SP --set notifyMode=mentions. any search defaults to hybrid and degrades to FTS when the server has no embedder — the reply's mode says which ran; --limit counts records, --max-data bounds each hit's text window (default 512 runes, -1 = whole chunk), --passages adds up to 10 further matching chunks per record, --filter keeps hits whose host object matches an objects-query filter (inline JSON, @FILE or -).

Space create and list have no dedicated subcommand — use curl against POST /v1/spaces and GET /v1/spaces; join with any join --token T below.

One-to-one spaces

any one-to-one start    <otherIdentity>              # aliases: any 1-1, any direct
any one-to-one accept   <spaceId>
any one-to-one decline  <spaceId>                    # sticky; a later start un-declines
any one-to-one register <peerIdentity> [--name N] [--description D] [--icon-cid CID]
any one-to-one pending                               # incoming requests

Catalog and bundles

any catalog list                                     # GET /v1/catalog
any catalog get   <usecaseId>
any catalog setup <usecaseId> <spaceId>              # adopt-or-install, dependencies first

any bundle ensure  <spaceId> --body '<json>'|@FILE|-
any bundle list    <spaceId>
any bundle get     <spaceId> <bundleId>
any bundle resolve <spaceId> <bundleId> <loserRootId>
any bundle child   <spaceId> <bundleId> --seed SEED --type T [--collection C]...

catalog setup prints every bundle it touched with its typeId or collectionId — whichever kind the bundle declares — and its xKey → propId properties, so the ids a client writes with come from the reply, never from minting a definition by xKey.

any catalog setup wiki $SP

Reads and writes

any query-subscribe <spaceId> <objectId> --dataset NAME [--filter J] [--sort K] [--limit N] [--offset N] [--total] [--projection P]
any query-subscribe <spaceId> --properties [same flags]      # per-space objects storage collection
any aggregate <spaceId> <objectId> --dataset NAME --pipeline '<json>'|@FILE|-
any aggregate <spaceId> --properties --pipeline '<json>'|@FILE|-
              [--group-limit N] [--accum-limit N] [--memory-limit N] [--explain]
any upsert <spaceId> <objectId> --dataset NAME --records '<json>'|@FILE|- [--page-size N] [--trace-id T]...
any backlinks <spaceId> <objectId> [--record R --dataset D | --prop P] [--kind K]... [--limit N]
any backlinks --target <any://uri> [--kind K]... [--limit N]  # across every indexed space
any links     <spaceId> <objectId> [--record R] [--dataset D] [--prop P] [--kind K]... [--limit N]

query-subscribe prints the ready → snapshot → changes → closed frames as JSON lines; --limit is required when --sort is set. --kind takes mention, link, card, embed or relation.

any query-subscribe $SP $CHAT --dataset chat_messages --sort=-_ver.id --limit 50 \
  | jq 'select(.event=="changes") | .data[]'

Snapshot query, modify and delete-records have no subcommand — call POST /v1/spaces/:id/query, /modify, /delete-records directly.

Editor

any editor blocks create <spaceId> <objectId> --type T [--text S] [--style JSON] [--parent ID] [--pos LEXID] [--collection NAME]
any editor blocks patch  <spaceId> <objectId> <blockId> [--set JSON] [--unset PATH]... [--collection NAME]
any editor blocks delete <spaceId> <objectId> <blockId> [--collection NAME]
any editor edit          <spaceId> <objectId> --old TEXT --new TEXT [--all] [--collection NAME]
any editor edit          <spaceId> <objectId> --edits '<json>'|@FILE|- [--collection NAME]

--collection names the editor collection (default editor_blocks; a namespaced <typeId>_<key> for a part with its own editor). editor edit is PATCH …/editor/:collection/markdown: each oldText must match the current rendering exactly (whole-line fuzzy fallback for unicode punctuation and trailing whitespace) and, without --all, exactly once. The example needs - [ ] buy milk in the document; once it applies, running it again answers markdown.no_match.

any editor edit $SP $DOC --old '- [ ] buy milk' --new '- [x] buy milk'

Chat

any chat send   <spaceId> <objectId> --text S | --file FILE|-  [--reply-to MSG]
                [--agent-name N] [--agent-debug-link L] [--agent-done=false]
any chat edit   <spaceId> <objectId> <msgId> --text S | --file FILE|-
any chat delete <spaceId> <objectId> <msgId>
any chat react  <spaceId> <objectId> <msgId> <emoji>          # toggle

<objectId> is the rootId of the space's chat — any catalog setup general-chat <spaceId> installs or adopts it and prints the root. Text is markdown; --file - reads it from stdin. Edit and delete work on your own messages only. Reading is any query-subscribe … --dataset chat_messages.

Objects

any object create <spaceId> --type T [--collection C]... [--properties '<json>'|@FILE|-]

any object type set          <spaceId> <objectId> <typeId>
any object collection attach <spaceId> <objectId> <collectionId>
any object collection detach <spaceId> <objectId> <collectionId>

--type is required (page is the plain document) and object type set replaces the previous one — there is no unset. --collection repeats; attach and detach are idempotent, and detaching is not a delete: that owner's values stay on the row as orphan data. --properties seeds values keyed owner → propId → value, the owner being the type or one of the collections.

set and attach pre-flight the id (404 type.not_found / 404 collection.not_found, 400 type.not_a_type / 400 collection.not_a_collection when it names the other surface); detach pre-flights nothing, so it can repair a row that already carries a bogus id.

any object create $SP --type page --collection $WIKI
any object create $SP --type $PERSON --collection $CONTACT --properties '{"any":{"name":"Ada"}}'
any object collection attach $SP $OBJ bin     # to the bin
any object collection detach $SP $OBJ bin     # restore

Types and properties

any type create <spaceId> --name N --xkey K [--description D] [--icon-cid CID]
                [--layout '<json>'] [--hidden] [--meta k=v]...
any type list   <spaceId> [--include-hidden]
any type update <spaceId> <typeId> [--name N] [--description D] [--icon CID]
                [--layout '<json>'|''] [--hidden[=false]] [--meta k=v]...

any type property list   <spaceId> <typeId>                        # alias: any type prop
any type property add    <spaceId> <typeId> --name N --kind string|number|boolean|array|object|datetime
                         [--xkey K] [--description D] [--scope synced|account|local] [--x-format '<json>'|@FILE|-]
any type property patch  <spaceId> <typeId> <propId> --set '<json>' [--unset PATH]...
any type property remove <spaceId> <typeId> <propId>
any type property option set    <spaceId> <typeId> <propId> <key> [--name N] [--color C] [--pos LEXID]
any type property option delete <spaceId> <typeId> <propId> <key>

any type part list   <spaceId> <typeId>
any type part add    <spaceId> <typeId> --draft '<json>'|@FILE|-
any type part patch  <spaceId> <typeId> <partId> --set '<json>' [--unset PATH]...
any type part remove <spaceId> <typeId> <partId>

any type part dataset list   <spaceId> <typeId>                    # alias: any type part ds
any type part dataset add    <spaceId> <typeId> <partId> --draft '<json>'|@FILE|-
any type part dataset patch  <spaceId> <typeId> <defId> --set '<json>' [--unset PATH]...
any type part dataset remove <spaceId> <typeId> <defId>
any type part dataset field add    <spaceId> <typeId> <defId> --field '<json>'|@FILE|-
any type part dataset field patch  <spaceId> <typeId> <defId> <fieldId> --set '<json>' [--unset PATH]...
any type part dataset field remove <spaceId> <typeId> <defId> <fieldId>

any type property add $SP $T --name Stage --xkey stage --kind array --x-format '{"type":"choice"}'
any type property patch $SP $T $P --set '{"xFormat.options.high.name":"High","xFormat.options.high.color":"red"}'
any type property option set $SP $T $P high --name High --color red --pos a0    # same write, sugar

Collections

any collection list   <spaceId> [--include-hidden]
any collection get    <spaceId> <collectionId>
any collection create <spaceId> --name N --xkey K [--description D] [--icon-cid CID]
                      [--hidden] [--meta k=v]...
any collection update <spaceId> <collectionId> [--name N] [--description D] [--icon CID]
                      [--hidden[=false]] [--meta k=v]...

any collection property list   <spaceId> <collectionId>            # alias: any collection prop
any collection property add    <spaceId> <collectionId> --name N --kind string|number|boolean|array|object|datetime
                               [--xkey K] [--description D] [--scope synced|account|local] [--x-format '<json>'|@FILE|-]
any collection property patch  <spaceId> <collectionId> <propId> --set '<json>' [--unset PATH]...
any collection property remove <spaceId> <collectionId> <propId>

A collection is what an object is filed under; a type is what it is. A collection takes a name, description, icon, --xkey, --hidden, --meta and property definitions — no parts, no layout. The property verbs are the type ones on a collection owner, same flags and same {set, unset} patch rules; the choice-option sugar is on the type group only, so patch the descriptor path directly here.

--xkey is required on create and unique across the space's types and collections together (409 type.xkey_conflict). collection list omits hidden collections (the built-in miniapp / bin, and any you marked hidden) without --include-hidden; collection get resolves them always. A registered built-in refuses a metadata write (400 collection.registered), and deleting a collection is 501 sdk.not_implemented.

any collection create $SP --name Contacts --xkey contact
any collection property add $SP $C --name Company --xkey company --kind string
any collection update $SP $C --hidden=false --meta pinned=true

Files

any file attach   <spaceId> <objectId> <path>|-  [--name N] [--mime M] [--variant V --variant-of FILE]
any file list     <spaceId> [--object OBJ] [--limit N]
any file get      <spaceId> <fileId>
any file download <spaceId> <fileId> [-o PATH] [--variant V]     # raw bytes to stdout by default
any file status   <spaceId> <fileId>
any file stats    <spaceId>
any file subscribe <spaceId>
any file pin | retry | offload <spaceId> <fileId>
any file delete   <spaceId> <fileId> --yes
any file query    <spaceId> <objectId> [--filter J] [--sort K] [--limit N] [--offset N] [--total] [--projection P]
any file query-subscribe <spaceId> <objectId> [same flags]
any file cache size | free <bytes> | sweep

The attach receipt normally shows durable: false — backup is background work; any file subscribe shows the inflight → durable flip. offload exits 1 with file.not_durable while the local bytes are the only copy.

Local store

any local meta                                       # pipeline stages and accumulators
any local collections [--scope account|space] [--space ID]
any local ensure  <name> [--space ID] [--index 'a,-b']... [--unique-index 'k']...
any local drop    <name> --yes [--space ID]
any local insert | upsert <name> --doc '<json>'|@FILE|- [--space ID]
any local update  <name> <id> --modifier '<json>' [--upsert] [--space ID]
any local delete  <name> [<id>...] [--filter J] --yes [--space ID]
any local get     <name> <id> [--space ID]
any local query   <name> [--filter J] [--sort K] [--limit N] [--offset N] [--total] [--projection P] [--space ID]
any local aggregate <name> --pipeline '<json>' [--group-limit N] [--accum-limit N] [--memory-limit N] [--explain] [--space ID]
any local indexes <name> [--ensure 'a,-b']... [--unique-ensure 'k']... [--drop NAME]... [--space ID]
any local export  [--scope account|space] [--space ID] [--names a,b] --out FILE
any local import  FILE                               # - reads stdin

Without --space a local storage collection is account-scoped; with it, bound to that space. export with neither scope nor names takes every local collection. Nothing here syncs.

any local ensure scratch --index k,-at
any local insert scratch --doc '[{"id":"a","k":1},{"k":2}]'
any local query scratch --filter '{"k":{"$gt":0}}' --sort -k --total

With that collection on an authorized server, export and re-import it:

any local export --scope account --names scratch --out scratch.anyenc.gz &&
any local import scratch.anyenc.gz

--out - writes the export to stdout; any local import - reads stdin. Without --names, export takes every collection in the selected scope; --space ID implies space scope. The file is a gzip-compressed anyenc stream carrying scope, space id, names, indexes and documents. Another authorized server imports it with any --addr http://127.0.0.1:7002 local import scratch.anyenc.gz; a space-scoped collection is readable there even when that server never had the space.

Import ensures indexes and upserts documents; documents whose ids are absent from the file stay. It commits in chunks, so a failed import can leave earlier chunks in place. It copies local collections only — no CRDT data, no file bytes. Responses and failure codes: HTTP contract.

Members, invites, ACL

any members list | me | requests <spaceId>
any members get <spaceId> <identity>
any members subscribe <spaceId>

any invite create | list | revoke-all <spaceId>
any invite get | revoke <spaceId> <recordId>
any invite guest-key <spaceId>                      # public read-only token (owner)
any invite guest-key-revoke <spaceId>               # rotate the read key
any invite pending                                  # direct-add invites awaiting approval
any invite accept | decline <spaceId>

any join --token T [--name N] [--description D] [--icon-cid CID]

any acl accept       <spaceId> --record R [--permission P]          # R from `any members requests`; default writer
any acl decline      <spaceId> --identity X
any acl grant        <spaceId> <identity> <permission>
any acl remove       <spaceId> <identity>...
any acl add          <spaceId> <identity>[,<identity>...] <permission> [--name N] [--description D]
any acl ownership    <spaceId> --new-owner X [--old-owner-perm P]   # default admin
any acl self-remove | cancel-join | stop-sharing <spaceId>

Permissions are none, reader, guest, writer, admin, owner. any acl add adds accounts by identity in one ACL record; each added account sees an invite_pending row and resolves it with any invite accept / decline. --name / --description apply to a single-identity add only.

Sync status, events, processes, push, debug

any sync-status space     <spaceId>
any sync-status object    <spaceId> <objectId>
any sync-status subscribe [<spaceId> <objectId>]    # account-wide without args

any events publish --type T [--scope device|account|space] [--space ID] [--target X] [--data J|@FILE|-]
any events subscribe [--scope S]... [--type T|x.*]... [--space ID]... [--target X]...

any process list
any process cancel ID [--identity X]

any push token set --platform ios|android --token TOKEN
any push token revoke | status
any push subscriptions

any debug space  <spaceId>                          # per-peer headsync counters
any debug object <spaceId> <objectId>               # tree + sync snapshot (walks the tree)
any debug p2p                                       # LAN discovery snapshot

any events publish --type ui.open_space --data '{"spaceId":"'$SP'"}'
any events subscribe --type 'ui.*'

Note. GET …/sync-status/peers has no subcommand and returns 501 until the SDK exposes a stable per-space peer list; any debug space is the diagnostic equivalent. The push commands exit 1 with push.disabled unless the server has a push node configured.