Simplified CLI
Skill ID: simplified-cli · Install skills · Canonical source
Try it
Help me use smp to search assets in my chosen teamspace.
The instructions below describe the workflow. Installed skills and connected tools are separate; use only operations exposed by your authorized connection.
smp is supplied by the Python simplified-apikit package. It generates commands from the same OpenAPI specs and runs them through MCP middleware, so request validation and hooks apply across CLI and connector calls. It is different from the npm simplified CLI; do not mix their commands or credentials.
Setup and scope
Use your team’s supported package source to install simplified-apikit. Confirm smp --version and command help before scripting. Local CLI calls require configured credentials; hosted app OAuth does not automatically populate local CLI environment variables.
Set SMP_TOKEN privately in the environment. API keys use their bound workspace; DRF tokens also require numeric SMP_WORKSPACE. Set numeric SMP_SPACE only after resolving the user’s chosen teamspace. Omit SMP_URL for production; use overrides only for an explicitly selected environment. Never print credentials or put them into user-facing commands/logs.
Root options such as --space, --workspace, and --raw precede the namespace. Both smp api list-assets and smp api:list-assets work. JSON options take a single shell-quoted JSON value.
Use the same resolved scope throughout. Do not guess tenant IDs or select the first board/account/member just to make a script run.
Choose the namespace
Comments are smp api:list-comments / smp api:add-comment, not a separate comments namespace. See comments.
Execute and verify
Read the target state, resolve IDs, then execute the user-authorized operation. Use --help for the exact required flags. Keep returned IDs for follow-up; verify writes using direct reads. For partial batches, report successful IDs and failed steps without repeating successful creates.
Current image/video generation waits for completion. Pending/timeout responses require continuation using the actual job identifiers, not a duplicate paid request. Video submission task IDs do not establish render completion. Other tools may have different task-result contracts.
Default output is formatted JSON; root --raw makes it suitable for JSON parsers. Read the actual envelope before writing a jq/path expression. Pagination limits and defaults vary by endpoint; retain filters/scope and follow returned pagination metadata.
Generation consumes credits and social scheduling/queueing can publish content. Preserve explicit authorization already given for the concrete operation. Planning is not permission to run it.
Local MCP server
full is the default local profile. Hosted connector tools/profile are deployed separately; local full access does not prove a hosted client has the same tools. flows, agents, and the generic passthrough require an automation/full surface. See client setup.
