Plans and permissions:
- Plan: Pro or Enterprise plan
- Read Assembled data through MCP: Admin role, or a custom role with Assembled MCP set to Can access all data via the MCP or Can access and make changes to all data via MCP
- Create forecast adjustments and publish overtime or voluntary time off through MCP: Admin role, or a custom role with Assembled MCP set to Can access and make changes to all data via MCP
- MCP access gives a role read access to all data available through MCP, whatever the role’s other permissions or data restrictions.
Use the Assembled MCP server to ask an AI assistant, such as Claude or ChatGPT, questions about your schedules, forecasts, adherence, and performance in plain English, and get answers from your Assembled data without running a report or an export.
To connect your AI assistant, see Assembled MCP server setup. The server address is https://data.mcp.assembledhq.com/mcp.
MCP, the Model Context Protocol, is the standard AI assistants use to connect to other tools. Your assistant can also combine Assembled data with anything else it has access to, such as your HR system or BI tools.
Note: The Assembled MCP server is separate from MCP connectors, which let Assembled’s AI agents call tools in your other systems during customer conversations. Connecting the MCP server doesn’t turn on AI agents or give them access to your systems.
How the Assembled MCP server works
- You connect with your own Assembled account, by signing in to Assembled through OAuth when your AI assistant asks.
- MCP access isn’t limited to what a role can see in Assembled. A role with MCP access can read all of your company’s data that’s available through MCP, even if the role is limited to one team or site in Assembled.
- The connection uses the Assembled instance you signed in to. To use another instance, disconnect and sign in again with that instance’s account.
- MCP reads only Assembled’s workforce management data: schedules, forecasts, adherence, and performance metrics. It doesn’t reach any system outside Assembled.
- Data is available from October 2025 onward.
Important: Connect MCP only from your organization’s approved AI assistant, and treat MCP access like any other integration’s credentials. Anyone with MCP access can read all of your company’s data that’s available through it.
What you can do with MCP
Your AI assistant chooses the right tools for each question. You don’t need to know them, but they help when you set tool permissions.
| Area | What you can do | Tools |
|---|---|---|
| Metrics | Browse the metrics catalog, and query individual, team, queue, SLA, CSAT, and other operational metrics |
get_metrics_catalog, query_metrics
|
| Forecasts | Compare forecasted and actual volume, handle time, service level, and staffing. List and create forecast adjustments. |
get_forecasted_vs_actual, get_forecast_adjustments, create_forecast_adjustments
|
| Staffing | See staffing parameters, such as shrinkage, productivity, handle time, and service level targets for each queue, and compare required, scheduled, and actual heads for BPOs |
get_staffing_parameters, get_bpo_heatmap
|
| Scheduling | Read who’s scheduled when, see who changed a team member’s schedule and when, and total each person’s overtime, time off, and voluntary time off |
get_schedule, get_schedule_history, get_agent_time_summaries
|
| Overtime and voluntary time off | View overtime and voluntary time off slots, and publish new ones |
get_overtime_and_vto, create_overtime_and_vto
|
| Agent states | Read team members’ states from your connected platforms | get_agent_states |
| Setup | Find the IDs of queues, sites, teams, skills, and channels, and get the current date and time |
get_workforce_config, get_current_time
|
create_forecast_adjustments and create_overtime_and_vto change data in Assembled. They need the Can access and make changes to all data via MCP option, and most AI assistants ask you to approve each use.
Which metrics you can query depends on the metrics available to your company.
For tips on writing prompts, and example prompts to try, see Prompts for the Assembled MCP server.
What MCP doesn’t include
- AI agent results, such as chat, email, and voice AI metrics.
- The dates and types of pending time off requests. Time off totals can include pending requests if you ask.
- People and headcount data, such as team members’ skill history and queue assignments.
- Forecast and actual comparisons across all channels at once, over more than about 6 months, or for test schedules. Each comparison covers one channel and your primary schedule.
Security and compliance
- Your AI assistant runs the model, so its provider’s policies apply. Assembled doesn’t use your data to train models.
- Assembled logs operational details, such as which tool was called, how long it took, and whether it worked, but doesn’t store your prompts.
Assembled is SOC 2 Type II audited by an independent auditor, and meets GDPR, CCPA, and HIPAA requirements. Data is encrypted in transit and at rest. For details, see trust.assembled.com.
If your company’s IT policy doesn’t allow custom connectors in your AI assistant, your customer success manager can share security documentation for your IT team to review.