.AIAPI ReferenceAPIANT.aiv3
Skills
The procedures Claude loads over the APIANT MCP connection, grouped by workflow, with what each one lets you ask Claude to do.
A skill is a set of instructions Claude follows when your request matches it, so you ask in plain English and Claude picks the skill. Claude loads the skills over the APIANT MCP connection when it starts working on your account; they are not files in your project, and they update on APIANT's side. Some skills match requests you make directly; the others are loaded by Claude while it works on one of those.
This list is generated from the skill source and checked against the skills served over the connection on 2026-09-19. To connect, see Connect your AI client.
How to read the tables
| Column | Meaning |
|---|---|
| What you can ask Claude to do | The task the skill performs. |
| Example asks | Phrases taken from the skill's own trigger list. Any wording with the same meaning works. |
| Assistant | Yes means the in-app assistant also uses this skill, in a version without the parts that need your own AI client. The assistant has no connector, publish or alert-tuning skills. |
| Skill | The skill's name, for reference. Describe what you want rather than typing the name. |
Build and edit automations
| What you can ask Claude to do | Example asks | Assistant | Skill |
|---|---|---|---|
| Build a new automation from a plain-English description of the workflow. | "build an automation", "connect X to Y" | Yes | build-automation |
| Change an existing automation: add, move or disable steps, fix mappings, swap one app for another. | "add a step", "fix the mapping" | Yes | edit-automation |
| Apply the step comments you left on an automation diagram in the web app, across all your automations in one pass. | "apply my comments", "check the diagram" | Yes | apply-comments |
| Build a suite of several coordinated automations, such as a two-way sync or a data pipeline. | "build an integration", "build a data pipeline" | No | build-integration |
Automation patterns Claude loads
Claude loads these while building or editing an automation, when your request matches the pattern. You describe the outcome you want; you do not ask for them by name.
| What you can ask Claude to do | Example asks | Assistant | Skill |
|---|---|---|---|
| Hosted forms: run an automation on every submission of a public form, or pause a run until a one-time form is filled out. | (none) | Yes | pattern-forms |
| Chat widget automations, from a simple AI chatbot to multi-goal conversations that call tool automations. | (none) | Yes | pattern-chat-widget |
| Web service automations: an HTTP endpoint where the caller waits for the automation's response. | (none) | Yes | pattern-web-service |
| Automations exposed as callable tools to MCP clients. | (none) | Yes | pattern-mcp-tool |
| Human approval: pause a run until a person approves or denies an item from a moderation queue link. | (none) | Yes | pattern-human-moderation |
| Parent and child automations: one automation calls another, passes inputs, or fires several without waiting. | (none) | Yes | pattern-execute-automation |
| Fan-out and fan-in: launch several child automations in parallel and wait for all of them to finish. | (none) | Yes | pattern-latches |
| Pause a run until a future date and time. | (none) | Yes | pattern-snooze |
| Collect items into named buckets across runs, then process the bucket on a schedule or trigger, for example a daily digest. | (none) | Yes | pattern-collector |
| Import field mappings from a CSV file or spreadsheet you provide. | (none) | Yes | pattern-csv-mapping |
| Small one-action recipes: read or publish RSS and Atom feeds, generate PDFs. | (none) | Yes | pattern-system-utilities |
| Keep records in two apps in sync in both directions, using two-way sync triggers and actions. | (none) | Yes | assembly-bidirectional-sync |
Test
| What you can ask Claude to do | Example asks | Assistant | Skill |
|---|---|---|---|
| Run an automation, inspect the results, diagnose and fix failures, retest, and cover each branch. | "test this automation", "why is automation X failing" | Yes | test-automation |
| Test a multi-automation suite end to end: data flow between automations, shared state, and loop prevention. | "test the integration end to end" | No | test-integration |
Apps and connectors
A connector is the APIANT definition of an app's API: its authentication plus the triggers and actions built on it. Whether your account can create connectors depends on its permissions and plan; see Build a connector, action or trigger.
| What you can ask Claude to do | Example asks | Assistant | Skill |
|---|---|---|---|
| Build a connector, or add an action or trigger to an app, against the app's API. | "create a connector", "add an API action" | No | build-assembly |
| Change an existing action, trigger or connector: the API call, the endpoint, a setting, a dropdown. | "update the API call", "change the endpoint" | No | edit-assembly |
| Read an app's own API documentation and work out which actions and triggers its connector should have. Read-only: nothing is written. | "what ops should we build for X" | No | ops-discovery |
| Register an app on the vendor's developer portal and connect it. Claude drives the portal in a browser (the page, the form, the scopes, the redirect URI); you sign in and answer any MFA. Needs browser control in your Claude session. | "I need my own keys for this app", "get the client id and secret" | No | get-creds |
| Connector authentication: OAuth 2.0, OAuth 1.0a, API key, no credentials, token-mint, and request signing. Loaded by Claude, not asked for directly. | (none) | No | assembly-connector |
| Build an action that creates a record. Loaded by Claude, not asked for directly. | (none) | No | assembly-action-add |
| Build an action that updates a record. Loaded by Claude, not asked for directly. | (none) | No | assembly-action-update |
| Build an action that deletes a record. Loaded by Claude, not asked for directly. | (none) | No | assembly-action-delete |
| Build an action that fetches one record. Loaded by Claude, not asked for directly. | (none) | No | assembly-action-get |
| Build an action that searches for records. Loaded by Claude, not asked for directly. | (none) | No | assembly-action-find |
| Build an action that lists records. Loaded by Claude, not asked for directly. | (none) | No | assembly-action-list |
| Build an action that calls an operation returning computed data, such as AI inference or a transformation. Loaded by Claude, not asked for directly. | (none) | No | assembly-action-invoke |
| Build a polling trigger for new items. Loaded by Claude, not asked for directly. | (none) | No | assembly-trigger-new |
| Build a polling trigger for new or updated items. Loaded by Claude, not asked for directly. | (none) | No | assembly-trigger-updated |
| Build an instant trigger that registers its webhook with the vendor when the automation turns on and removes it when it turns off. Loaded by Claude, not asked for directly. | (none) | No | assembly-trigger-self-registering-webhook |
| Build a webhook trigger for an app that sends its events to a fixed URL (one per connected account, or one for the whole app) and has no API to register webhooks. Loaded by Claude, not asked for directly. | (none) | No | assembly-trigger-service-webhook |
| Build a webhook trigger whose URL you paste into the app's dashboard yourself. Loaded by Claude, not asked for directly. | (none) | No | assembly-trigger-manual-webhook |
Publish and deploy
Plan: Enterprise. See Publish and deploy for the permissions publishing and deploying need.
| What you can ask Claude to do | Example asks | Assistant | Skill |
|---|---|---|---|
| Publish template automations from dev to prod, then deploy them to customer accounts, linked or unlinked, chosen per deploy. | "push to prod", "update customers" | No | deploy-automation |
Alerts and errors
| What you can ask Claude to do | Example asks | Assistant | Skill |
|---|---|---|---|
| Inspect and tune how errors in your automations are handled: why a run did or did not retry, why an automation was turned off after a 401, how an alert message was rewritten, and which alerts are suppressed. The platform-wide retry and 401 lists are read-only from your account; Claude gathers the evidence for APIANT support instead. | "why didn't my automation retry", "noisy alerts" | No | alert-handling |
Feedback
| What you can ask Claude to do | Example asks | Assistant | Skill |
|---|---|---|---|
| Send a bug report or suggestion to APIANT. When you ask, Claude drafts it in your words and files it only after you say yes. Claude also files a report on its own, without asking, when the APIANT MCP server returns broken or wrong responses. | "send feedback", "I have a suggestion" | Yes | report-mcp-issue |
Next steps
Related docs
Last updated September 19, 2026