Stacksync MCP Gateway
Give AI assistants controlled access to external apps through your existing Stacksync connections
Stacksync MCP Gateway lets your AI assistant read data and take actions in external systems through a controlled layer. It uses the apps connected to your Stacksync workspace, including CRMs, ERPs and databases, with access controls, guardrails and activity logging managed in Stacksync.
The gateway is powered by the same Stacksync MCP server you use to manage the Stacksync platform. Connect your AI assistant once: the same connection gives it both capabilities, subject to your permissions and workspace settings.
How it works
- You connect your AI assistant to Stacksync and authenticate.
- Your assistant finds the connected app and the action needed for your request, and fills in the action's form with live values from the app.
- Stacksync checks the applicable MCP Gateway Policies for access to the app and action. Changes also pass through the policy's guardrails and any required approval.
- An allowed action runs through the workspace's existing connection. Its outcome appears in the MCP Activity Log.
You can run an app action directly without building a workflow. Credentials stay in Stacksync. To connect a new app, your assistant gives you a link to connect it in the Stacksync app rather than asking for credentials in chat.
Example requests
- "Find the HubSpot contact for jane@acme.com and set her lifecycle stage to Customer."
- "Search Salesforce for the Acme account and show its open opportunities."
- "Read the orders table in PostgreSQL and show the latest ten orders."
The available actions depend on the app, its connection and your access.
Access controls and custom guardrails
Workspace owners use MCP Gateway Policies to choose which apps and actions users and groups can access. Every workspace has a Default Policy; custom policies provide access for specific teams.
Guardrails add rules in plain language for changes through an app, such as limiting purchase orders to a subsidiary or restricting which fields can be updated. Stacksync AI checks proposed changes against those rules. It blocks changes that violate them and requests confirmation when the result is uncertain. Read operations are not subject to guardrail checks.
Configure MCP Gateway Policies and guardrails.
Logging and monitoring
The MCP Activity Log shows calls, blocked actions, approval requests and errors, with the member, assistant, resource and outcome. Owners can filter activity across the workspace; other members see their own activity. It records call metadata, not the data an assistant reads or sends, or credentials.
Governance and security
Your assistant acts on your behalf with your workspace permissions. Owners can require approval with Ask before every change and control access to troubleshooting logs. You can revoke an assistant's session from the MCP page.