What is MCP Server Console?
MCP Server Console is the inbound surface of ServiceNow's Action Fabric, giving enterprises a centralized workspace to register, govern, and monitor the AI tools they expose to external agents and clients. Every tool published through the console follows the Model Context Protocol (MCP) standard, making it discoverable and callable by any compliant AI agent.
MCP Server Console 1.8 delivers on the Observe pillar with a new Monitoring workspace that surfaces live tool call activity directly in the console. This release also introduces safe customization of ServiceNow-delivered tools through cloning with lineage tracking, and two new tool types: AI Search, which grounds agent responses in enterprise knowledge, and Playbooks, which gives AI agents the ability to trigger and execute structured ServiceNow workflows.
WHAT SHIPPED IN 1.7
Previous release highlights
|
1
Connect AI Clients in Minutes
A unified onboarding wizard lets admins register any MCP client, copy all required credentials from a single screen, and go live in minutes rather than coordinating across multiple OAuth configuration pages.
|
2
Control What Agents See
Tool-level visibility settings give admins fine-grained control over which tools are discoverable by which AI agents, enforcing least-privilege access without removing the tool from the server entirely.
|
3
Label AI Tools by Risk Level
Standardized safety annotations, including read-only and destructive hints, signal to AI agents which tool calls carry higher risk, enabling compliant behavior and satisfying enterprise AI safety requirements.
|
WHAT'S NEW IN 1.8
MCP Server Console 1.8
For more information, see the MCP Server Console release notes and product documentation.
1 Monitoring Dashboard NEW
The Monitoring Dashboard is now available in MCP Server Console under Configuration > Monitoring. It shows tool call volumes and outcomes for a selected time window, grouped by tool or by server. The sn_mcp_server.viewer role is required.
|
BEFORE
Investigating tool call failures meant querying system logs. There was no centralized view of call volumes, success rates, or failure types across MCP servers.
|
NOW
Open Configuration > Monitoring, set a time window, and choose a Tools or Servers view. Call volumes, success rates, and failure breakdowns are all in one table. Enable per-server detail to see exactly which servers contributed to failures on a specific tool.
|
Two view modes: Tools and Servers
The Aggregated by control sets the grouping key for all rows in the table:
- Tools - One row per tool. The Server column lists every server that tool ran on. Both views show the same seven outcome columns.
- Servers - One row per server, showing which tools ran on it and how each performed.
In Tools view, toggle Show per-server detail to expand any tool row into individual server rows. A Total aggregate row appears at the top of each expanded group. Use this to identify which specific server is driving failures on a tool when the aggregate row shows a problem.
MCP Server Console Monitoring Dashboard in Tools view showing call volumes and outcome columns across all tools
Time window
Select Last 15 min, Last hour, or Last 24 hour from the Time window control, then click Apply filters. The label above the table confirms the active window and the total count of tools or servers in scope. Your selections persist when you switch between Tools and Servers views, so you do not need to reapply them.
Outcome columns
Each row reports seven columns. Failed is a total, not a separate category. Throttled, Denied, and Other errors are already counted inside Failed. Do not add them to the Failed number.
- Calls - Total call attempts in the selected window
- Success - Calls that completed successfully
- Failed - Total failed calls. Includes Throttled, Denied, and Other errors.
- Success rate - Success divided by Calls, shown as a percentage with a visual bar
- Throttled - Calls rate-limited by the server or platform. Already included in Failed.
- Denied - Calls blocked due to authorization or policy. Already included in Failed.
- Other errors - Failures from errors or exceptions not classified as throttled or denied. Already included in Failed.
Starting an investigation
|
1
|
Find the problem tool
Open the Tools view with Last 15 min selected and click Apply filters. Scan the Success rate column for any tool with a low or missing bar. Tools sorted near the top will have the highest call volumes, so low success rates there have the widest blast radius.
|
|
2
|
Isolate to server or tool
Enable Show per-server detail on the affected tool. If failures are concentrated on one server and other servers for the same tool are healthy, the issue is server-specific: check its configuration, rate limit settings, or recent changes. If all servers for that tool are failing, the issue is in the tool itself or its underlying artifact.
|
|
3
|
Diagnose by failure type
Check the Denied, Throttled, and Other errors columns separately. Each has a different cause and a different next step:
- Throttled - The tool or server is receiving more calls than its configured rate limit allows. Review the rate limit settings on the affected MCP server.
- Denied - Calls are being blocked at the authorization or policy layer. Check the tool's visibility settings in MCP Server Console and confirm the connecting client has the correct permissions to invoke the tool.
- Other errors - A runtime error occurred during tool execution. Review system logs for error details associated with the affected tool to identify the specific cause.
|
Choosing a time window
- Last 15 min - Use for active triage. Something is failing right now and you need current state.
- Last hour - Use after a configuration change. Confirm the change had the expected effect on call outcomes.
- Last 24 hours - Use for pattern analysis. Identify recurring failures or understand baseline call volumes before making configuration decisions.
Monitoring Dashboard with Show per-server detail enabled, look_up_incident_records expanded across 10 servers
Who this affects: Platform Admin, AI Steward, Compliance & Security
2 Safe Customization of Platform Tools NEW
Admins can now clone any ServiceNow-delivered MCP tool into a customer-owned, fully editable copy. The cloned tool maintains lineage back to its source in the base system, so when ServiceNow ships an update to the original, the delta is visible and can be applied to the custom copy without overwriting existing modifications. This removes the risk of upgrade-breaking changes caused by direct edits to out-of-the-box tools.
Upgrade-safe lineage tracking
Cloned tools stay linked to their source in the base system. When the platform-delivered version updates, the lineage relationship surfaces the change, so it can be selectively applied to the custom copy rather than requiring a full rebuild from scratch.
Tool cloning interface in MCP Server Console showing the clone action and lineage reference field
Who this affects: Platform Admin, Agent Builder
3 AI Search as a Tool NEW
Admins can now create an AI Search tool in MCP Server Console, giving AI agents the ability to query enterprise knowledge directly through a configured search profile. The search profile defines the scope of content the tool can access. The agent never reaches outside that boundary.
This connects AI agents to the verified, governed content your teams have already built, rather than relying on generated answers from model training data alone.
Search profile as the governance boundary
When creating the tool, you select a search profile. That selection defines exactly what content this agent is permitted to search: which knowledge bases, which catalog items, which content sources. Different tools can reference different profiles, allowing you to scope search access by use case or audience.
AI Search tool creation in MCP Server Console showing search profile selection
Who this affects: Platform Admin, Agent Builder
4 Playbooks as a Tool NEW
Platform administrators can now expose a ServiceNow playbook as an MCP tool, enabling any connected AI agent to trigger and execute the playbook through MCP. When an agent calls the tool, the playbook runs activity by activity, prompting the agent for required inputs at each step and advancing through the workflow automatically. This brings structured, governed ServiceNow processes directly into AI-driven conversations and agentic use cases. See the product documentation for more information.
Playbook compatibility requirements
Before a playbook can be exposed as an MCP tool, it must meet the following requirements. All dependencies are scanned and validated before the playbook can be added as a tool.
- Activity types - Only these activity types are supported: Create Record Form, Record Form, Questionnaire, Send Email, Knowledge, Instruction, and Decisions.
- Subflows and actions - Any subflows or actions referenced in the playbook must be marked as compatible for MCP execution. If any dependency is not marked, the entire playbook is ineligible.
- Activity start - All activities and lanes must start immediately. Delayed start is not supported.
- Field types - Primitive field types only. Reference fields, attachment fields, rich text editors, and custom field types are not supported.
Playbook tool creation in MCP Server Console showing playbook selection
Who this affects: Platform Admin, Agent Builder, Playbook Author
FEATURE REFERENCE
What ships when
| Feature |
Roles |
Release |
| Connect AI Clients in Minutes |
Platform Admin, AI Steward |
1.7 |
| Control What Agents See |
Platform Admin, AI Steward, CISO |
1.7 |
| Label AI Tools by Risk Level |
AI Steward, CISO, Agent Builder |
1.7 |
| Monitoring Dashboard |
Platform Admin, AI Steward, CISO |
1.8 |
| Safe Customization of Platform Tools |
Platform Admin, Agent Builder |
1.8 |
| AI Search as a Tool |
Platform Admin, Agent Builder, Knowledge Manager |
1.8 |
| Playbooks as a Tool |
Platform Admin, Agent Builder, Playbook Author |
1.8 |
