Connect An MCP Client

The Automation Suite is the local part of Daeda AI.

It runs as an MCP server (@daeda/mcp-pro) on your machine and syncs portal data into a local DuckDB database. Your AI client then reads that local database through MCP tools and proposes Write Plans for review inside Daeda AI.

Before You Start

Requirement	Why it matters
A HubSpot portal with Daeda AI installed	Daeda AI is the admin surface for the Automation Suite
An MCP license key from Daeda AI	The local client uses it to register
Node.js 20 or newer	@daeda/mcp-pro requires Node 20+
An MCP-compatible client	Claude Code, Cursor, and other MCP clients work

What You Set Up In HubSpot First

1. Open Daeda AI

Go to the Daeda AI settings page in HubSpot.

2. Create An MCP License Key

Open Master Settings.

Create a license key in the MCP License Keys section.

3. Connect Any Client Portals You Need

If you want to query a different HubSpot portal, connect it first from Master Settings.

Select the data scopes you want.

Generate the install link.

Open that link on the target portal and approve the scopes.

Scope Planning

Choose scopes based on the work you want the Automation Suite to do.

Scope type	Needed for
Read scopes	Querying CRM objects and metadata
Write scopes	Write Plans that create, update, delete, or merge records
Schema write scopes	Property, property-group, and custom-schema changes
Workflow and sequence scopes	Workflow metadata, triggers, and automation-related operations

If you request too few scopes, reads may be partial and some Write Plans will pause for re-authorisation.

Install In Claude Code

Run:

claude mcp add --transport stdio --env DAEDA_LICENSE_KEY=your-license-key mcp-pro -- npx -y @daeda/mcp-pro

Replace your-license-key with the key from Daeda AI.

Install In Other MCP Clients

Use the same package:

Setting	Value
Command	npx
Args	-y @daeda/mcp-pro
Env var	DAEDA_LICENSE_KEY=your-license-key

Example JSON:

{
  "mcpServers": {
    "mcp-pro": {
      "command": "npx",
      "args": ["-y", "@daeda/mcp-pro"],
      "env": {
        "DAEDA_LICENSE_KEY": "your-license-key"
      }
    }
  }
}

What Happens On First Connection

Step	What happens
1	The local MCP client starts over stdio
2	It registers with your license key
3	The server returns the connected HubSpot portals and the local encryption key
4	If this is the first real use and no paid plan is active, the free trial can start now
5	The client creates a local per-portal DuckDB database
6	It requests the latest artifact inventory for the selected portal
7	It downloads and ingests artifact-backed data
8	It starts live watch behaviour for the selected portal

Local Storage

By default, local portal data is stored under:

Path	Purpose
~/.daeda-mcp/portals/<portalId>/hubspot.duckdb	Live local DuckDB database
~/.daeda-mcp/portals/<portalId>/hubspot.replica.duckdb	Published read replica
~/.daeda-mcp/portals/<portalId>/portal_data.json	Local artifact and plugin sync state

First Verification

After the MCP client connects, verify in this order.

1. Confirm The Server Is Connected

In Claude Code, run /mcp.

You should see mcp-pro connected.

2. Check Connection State

Ask your AI client to call:

status(section="connection")

Look for:

Field	What you want
connectionState	Connected
selectedPortalId	A real portal ID
portals	The portals you expect to see

3. Select The Portal

If you have more than one connected portal, call:

set_portal(portalId=<your portal id>)

If only one portal is connected, it is usually auto-selected.

4. Check Schema Status

Call:

status(section="schema")

This is the main health check.

It shows the local database structure, table availability, plugin freshness, and refresh jobs.

5. Run A Simple Query

Call:

query(sql="SELECT id, json_extract_string(properties, '$.firstname') AS firstname FROM contacts LIMIT 5")

If your portal has contacts and the contact scope was granted, this should return rows.

The Most Important Tools

Tool	What it does
set_portal	Choose the default portal
status	Inspect connection, schema, freshness, and metadata
query	Run read-only SQL against the local DuckDB mirror
chart	Turn query results into a chart
configure_sync_controls	Enable or disable sync by portal or artifact
refresh_plugins	Refresh lightweight plugin-backed metadata
describe_operations	Discover write operation types
build_plan	Draft a Write Plan
submit_plan	Validate or submit a Write Plan
get_plans	List prior Write Plans

Multi-Portal Behaviour

The Automation Suite supports more than one connected client portal.

Situation	What to do
One connected portal	It is usually auto-selected
Multiple connected portals	Use set_portal before querying
Need to query many portals	Use tool calls with explicit portalIds

Sync Controls

Portal sync can be enabled or disabled per portal.

Artifact sync can also be disabled for specific object types.

State	Meaning
Portal sync enabled	The Automation Suite keeps artifact-backed data fresh for that portal
Portal sync disabled	Reads use stale local data only, if local data already exists
Artifact disabled	That object type stops receiving new artifact work

How Freshness Works

Two data systems exist in the local database.

Data type	Freshness model
Artifact-backed data	Managed automatically for normal reads
Lightweight plugin-backed data	Not real-time; refresh it manually when current values matter

Read this next:

How The Local Read Database Works

Common First-Time Issues

Problem	Likely cause	Fix
mcp-pro does not connect	Node version or bad config	Check Node 20+, command, args, and env
Invalid license key	Wrong key or revoked key	Generate a fresh key in Daeda AI
No portals returned	No connected client portals	Connect a portal in Daeda AI
Query says no data yet	Initial artifact sync still running	Wait, then re-run status(section="schema")
Query returns stale metadata	Plugin-backed table was not refreshed	Run refresh_plugins, then poll status(section="schema")
Registration blocked	Trial expired and no paid plan	Choose a paid plan in Daeda AI

Team Setup

For shared project configs, store the MCP config in the repo but keep the key in an environment variable.

Example:

{
  "mcpServers": {
    "mcp-pro": {
      "command": "npx",
      "args": ["-y", "@daeda/mcp-pro"],
      "env": {
        "DAEDA_LICENSE_KEY": "${DAEDA_LICENSE_KEY}"
      }
    }
  }
}

Next Steps

Next guide	Use it for
How The Local Read Database Works	Understand artifacts, plugins, freshness, and refresh rules
Troubleshooting	Diagnose setup, sync, and refresh issues
Custom Objects	The supported workflow for creating and using custom objects