Open Agent-Native apps beside the conversation through Dispatch MCP, a connection for app tools. Create or edit content using the current screen and selection, then read saved results back to confirm the changes.
Open and operate Agent-Native workspace apps through Dispatch MCP, with inline app surfaces, browser-session authentication, and current screen state.
Run score Not tested
See how to installReview the instructions the agent will receive before you install.
Use /an <app> to open an Agent-Native app beside the conversation and keep
the app's live UI and the agent's actions in the same workflow. The first
argument is the Dispatch app id. For example, /an slides opens the Slides
editor. After it opens, the user can sign in in that browser surface and ask
for new decks or focused edits in ordinary language.
For a known app, call Dispatch's open_app with an inline surface:
{
"app": "slides",
"path": "/",
"embed": true,
"chrome": "full"
}
The MCP App response is the preferred handoff for Claude, Claude Desktop, Claude Cowork, ChatGPT, ChatGPT desktop, Codex, and other compatible hosts. Keep the returned open link as the fallback for a host that cannot render MCP Apps inline. Do not claim that a browser opened unless the host rendered the inline surface or its browser tool confirmed the returned URL.
If the app name is missing or uncertain, call list_apps first and use the
exact granted app id. If the requested app is not listed, explain that it must
be granted in Dispatch's Agents access settings. Do not invent a URL, use a
localhost URL, or route around the Dispatch grant.
Prefer direct app tools when the host exposes them. After opening an app, call
list-host-webmcp-tools when available, then use run-host-webmcp-tool
with the exact listed name and origin. Page-local tools reflect current app
state and selection, so use view-screen before selection-dependent edits
and read back writes.
You are the model doing the work: when the user asks for new content — a design, a deck, a form, a document — author it yourself and save it with the app's create/update tools. Never hand off authoring to the app's own agent when those tools exist, and never wait on an in-app question form; it answers back through that app's own chat, not to you.
Use ask_app only when the host has no direct page tools, the requested
capability is not exposed, the task needs the app agent's interpretation or
multi-step specialist reasoning, or a direct call fails and needs recovery.
Dispatch's unified /mcp endpoint still exposes generic cross-app verbs; a
direct app MCP connection or page WebMCP surface is where named app actions
appear.
The app's MCP instructions name its key tools (generated from the app's own
list); follow those names verbatim and use tool-search for anything else —
never guess a tool name.
view-screen before any selection-dependent edit and use the ids it
returns; never guess an id from a title, index, or stale context.view-screen returns a selection and nextRequiredAction, make
that call directly with the smallest atomic change; do not regenerate a
whole artifact for a focused edit.ask_app_status task is not proof an edit completed.The Dispatch MCP connector identifies the calling agent, while the embedded app uses the user's browser session. Let the user complete sign-in in the rendered Agent-Native browser surface. Never request, copy, or store passwords, OAuth codes, cookies, or access tokens.
The generic Agent-Native plugin registers Dispatch at
https://dispatch.agent-native.com/mcp for hosts that support plugin MCP
configuration. ChatGPT custom connectors use the same URL with OAuth. After
installing or changing a connector, reload Claude/Cowork/ChatGPT or reconnect
and rescan the MCP server if the new tools or inline surface are not visible.
open_app succeeds but no inline surface appears, show its returned
link and say that this host does not expose an embedded browser here.list-host-webmcp-tools before delegating. If
it is available, use the exact returned tool name and origin for direct work.ask_app cannot reach Slides, check that Slides is granted in
Dispatch and that the connector is authenticated, then retry once with the
exact app id slides.No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Generate, refine, export, and place brand images or videos from an Assets library into another app. The skill covers choosing a library, handling image batches and video runs, and updating selected assets after feedback.
Prepare a customer account health, renewal, or contract usage readout. Verify the account identity, use approved metric definitions, compare contracted capacity with actual usage, and trace findings to sources, time windows, and known gaps.
Create focused app actions that agents can call as tools and frontend code can use through client hooks. Use them to add provider integrations or fetch UI data, with validated inputs and structured results.
Use this checklist when adding a feature or integration to an agent-native app. It covers the user interface, agent-callable actions, instructions, and app-state sync so both the user and agent can work with the capability consistently.
Investigate complex questions by combining data from multiple sources, then summarize patterns, metrics, and evidence in a clear report. Save requested analyses as dashboards with instructions for future refreshes.
Use the Analytics admin surfaces to inspect and change workspace feature flags, review dashboard ownership and usage activity, or browse and edit databases connected from other agent-native apps. The skill guides administrators to the relevant view and actions.