AnyCap CLI Reference

Use this reference to find command groups and inspect the inputs accepted by your installed CLI. If you are setting up AnyCap for the first time, follow installation and authentication first.

Syntax

anycap [global-flags] <command> [subcommand] [flags]

Global flags

Flag	Description
--endpoint <url>	Override the AnyCap server endpoint
--verbose	Write debug diagnostics to stderr
-v, --version	Print the CLI version
-h, --help	Show command help

Endpoint precedence is --endpoint, ANYCAP_API_ENDPOINT, config file, then the built-in production default.

Command groups

Group	Commands
Generation	image, video, music, audio
Understanding	actions
Web	search, crawl, download
Collaboration	annotate, draw
Delivery	drive, page, snapshot
Agent connectivity	mcp, connect, agent
Runtime	login, logout, credentials, status, config, update
Maintenance	skill, feedback, completion, help

Run help at any level:

anycap --help
anycap image --help
anycap image generate --help

Model discovery

Generation groups use the same discovery shape:

anycap <capability> models
anycap <capability> models <model-id>
anycap <capability> models <model-id> schema --mode <mode>

Treat the live catalog as authoritative. Static model lists become stale as availability and schemas change.

Example: discover a model and generate an image

List available models first:

anycap image models

Replace MODEL_ID with an ID returned by that command. Inspect the schema for the mode you intend to use:

anycap image models MODEL_ID schema --mode text-to-image
anycap image generate \
  --model MODEL_ID \
  --prompt "a paper crane on a wooden table" \
  -o paper-crane.png

On success, the image is downloaded to the specified output path. Use repeatable --param key=value flags for additional inputs defined by that model’s schema. Do not copy parameters from a different model without checking its contract.

For a complete workflow, see generate, refine, and share an image. For video inputs, follow image to video.

Output streams

Normal commands reserve stdout for one final JSON document. Invocation-started events and verbose diagnostics go to stderr. The mcp command reserves stdout exclusively for MCP protocol messages.

For a generation command, capture stdout and stderr separately so diagnostics do not corrupt the result your script parses:

anycap image generate \
  --model MODEL_ID \
  --prompt "a paper crane on a wooden table" \
  -o paper-crane.png > result.json 2> diagnostics.log

Check the command’s exit status before treating result.json as a successful generation. Help and version commands print human-readable text; anycap mcp is a long-running protocol server rather than a command that returns one result document.

Diagnose a failed command

Problem	Next step
Command or flag is not recognized	Inspect anycap --version and the command’s --help; update the CLI if the required command is missing.
Authentication is required	Run anycap login, complete authorization, and check anycap status.
A model or parameter is rejected	Refresh the model list and inspect the schema for the selected operation and mode.
More diagnostic detail is needed	Add --verbose and inspect stderr. Review logs before sharing them.

For MCP host configuration and file-access errors, use the MCP setup guide.

Shell completion

anycap completion --help

Follow the generated shell-specific instructions rather than sourcing output into an unsupported shell.