| name | zapier-status |
| description | Check the health of your Zapier MCP setup. Three modes — health check (dashboard view), audit (find waste and duplicates), diagnose (systematic troubleshooting). Use when asking "is my MCP working?", "check my tools", "audit my setup", "what's broken?", or "zapier status". |
Zapier status
Three modes for monitoring and maintaining a Zapier MCP setup. Determine the mode from context, or ask if unclear.
First, detect the server mode: If list_enabled_zapier_actions is available, the user is on Agentic mode. Otherwise, the user is on Classic mode where each configured action is its own MCP tool (e.g., slack_send_channel_message, gmail_find_email).
Mode 1: Health check
Trigger: "check my tools", "zapier status", "is everything working?", or any general status inquiry.
A quick dashboard view of the current state.
Steps
-
Check the available tools:
- Agentic mode: Call
list_enabled_zapier_actions to get the inventory of enabled actions.
- Classic mode: Inspect the available Zapier MCP tools. Each action tool follows the pattern
app_action_name. The built-in get_configuration_url tool is always present when the server is connected.
-
If no Zapier tools are available: report the connection status and suggest running zapier-setup (Classic) or calling get_zapier_skill with name "zapier-mcp-onboarding" (Agentic).
-
If tools/actions are available, build a summary by grouping actions by app (Agentic: from the list_enabled_zapier_actions response; Classic: from each tool's description):
For each app, show:
- App name
- Number of actions
- Action types (infer read vs write from names: find/search/get/list = read, send/create/update/add = write)
- Quick status: working / needs auth / error
Format as a dashboard:
Zapier MCP status
=================
Server: connected
Total actions: 14 across 5 apps
Slack 3 actions (2 read, 1 write) ✓ working
Gmail 3 actions (1 read, 2 write) ✓ working
Google Cal 2 actions (1 read, 1 write) ✓ working
Jira 3 actions (2 read, 1 write) ✓ working
Google Docs 2 actions (1 read, 1 write) ✓ working
-
If any actions appear broken (based on recent errors or auth issues), flag them.
-
End with: "Everything looks good." or "Found [N] issues. Want me to diagnose them?"
Mode 2: Audit
Trigger: "audit my setup", "clean up my tools", "find duplicates", "what should I remove?"
Find inefficiencies: duplicate actions, unused tools, conflicts with native MCP servers.
Steps
-
Get the full inventory:
- Agentic mode: Call
list_enabled_zapier_actions.
- Classic mode: Inspect the available Zapier MCP action tools.
-
Check for duplicates within Zapier MCP:
- Multiple actions for the same app that do similar things (e.g., both
slack_find_message and slack_search_messages)
- Recommend removing the less useful one
-
Check for conflicts with native MCP servers:
- Look at other MCP servers configured in the client's MCP config file (e.g.,
.cursor/mcp.json, claude_desktop_config.json, .mcp.json — depends on the client)
- If the user has both a native Slack MCP and Zapier Slack actions, flag it
- Recommend: "You have Slack through both Zapier and a native Slack server. The native server is usually better for single-app use. Consider removing Zapier's Slack actions."
-
Check for unused or low-value actions:
- Actions that are rarely useful as defaults (e.g., very specific write actions that are only needed occasionally)
- Suggest removing actions that can be re-added on demand through the web UI
-
Show the audit report:
Audit results
=============
Duplicates: 1 found
- Slack: "find_message" and "search_messages" overlap. Recommend removing "search_messages".
Native conflicts: 1 found
- Slack: native Slack MCP also configured. Consider removing Zapier Slack actions.
Cleanup candidates: 2 found
- Google Sheets "delete_row": rarely needed, can re-add on demand
- Jira "add_attachment": niche action, add when needed
Recommended removals: 4 actions
Want me to show you how to clean these up?
- If the user says yes:
- Agentic mode: Use
disable_zapier_action to remove the recommended actions directly in chat.
- Classic mode: Call
get_configuration_url and direct them to the web UI to remove the recommended actions. List exactly which ones to remove.
Mode 3: Diagnose
Trigger: "what's broken?", "my tools aren't working", "debug my MCP", or when a specific tool call has failed.
Systematic troubleshooting with error pattern matching.
Steps
-
Gather context: Ask what went wrong, or use the error from the current conversation.
-
Run diagnostics in order:
a. Connection check: Try calling get_configuration_url or any available Zapier tool. If nothing works, the problem is server-level (auth, config, network).
b. Action check:
- Agentic mode: Call
list_enabled_zapier_actions to see if the action is enabled. If not, use discover_zapier_actions to find it and enable_zapier_action to add it.
- Classic mode: Is the specific action tool available? If not, the user needs to add it through the web UI.
c. Auth check: Try calling a read action for the affected app (Agentic: execute_zapier_read_action; Classic: the specific read tool). If it returns an auth error, the app connection needs re-authentication.
d. Parameter check: Review the failing call's parameters. Common issues:
- Missing required fields
- Wrong field format (IDs vs names)
- Instructions that are too vague for the params resolver
-
Match against known error patterns:
| Symptom | Likely cause | Fix |
|---|
| All tools fail | Server auth expired | Re-authenticate at mcp.zapier.com |
| One app fails, others work | App-level auth expired | Re-connect that specific app |
| Tool not found / unavailable | Action not configured | Agentic: discover_zapier_actions + enable_zapier_action. Classic: direct user to get_configuration_url to add it |
| "invalid params" | Wrong fields or format | Check the tool's parameter schema |
| Results are empty but expected data exists | Search too narrow or wrong field | Broaden the search or check field names |
| Timeout on execute | Server overloaded or action is slow | Retry once, then report if persistent |
| "rate limit exceeded" | Too many calls | Space out requests, wait 30 seconds |
| Works in one project, fails in another | Project-level vs global config | Check both project-level and global MCP config for the client |
- Report findings:
"Here's what I found:
- Connection: OK (server responding)
- Actions: 14 tools available across 5 apps
- Auth issue: Gmail connection expired. You need to re-authenticate Gmail at mcp.zapier.com.
- Recommendation: [direct link or specific instruction]"
- If the fix requires user action (re-auth, config change), provide the specific link or instruction. If it's something the AI can fix (adjust parameters, try a different tool), offer to do it.
General notes
- Always check available Zapier MCP tools as the first diagnostic step in any mode. On Agentic, this means calling
list_enabled_zapier_actions. On Classic, this means inspecting the available tool names.
- Don't dump raw error messages. Translate them into plain language.
- If a problem is beyond what the skill can diagnose (server-side bug, API outage), say so and suggest checking status.zapier.com or contacting support.