Skip to main content
The Statisfy MCP server offers three sets of tools. Your AI assistant picks them based on what you ask, so you don’t call them yourself. A connection sees the tools its permissions allow: the tools that change data appear only on a connection allowed to make changes. See Read and write permissions.

Read your data

These tools mirror Statisfy’s data model, and only read:

Work across Statisfy

These tools reach the rest of Statisfy: reports, campaigns, email templates, surveys, knowledge bases, skills, Agent Studio, digital workers, workspace setup, and more. Each call runs as you, within your role. list_api_operations lists only the operations your role and your workspace’s products allow, so Claude isn’t offered one Statisfy would refuse.

Build your portals

If your workspace has Portals, the connector has tools to work on your customer portals. Reading a portal needs Portals access; editing one needs a role that can manage portals (Admin or Builder).

How a portal change reaches a preview

Your assistant follows a fixed path, and each step is a separate tool, so you can stop at any point:
  1. Pick the portal. It lists your portals and confirms which one you mean.
  2. Read. It lists the files and reads the ones it needs.
  3. Edit the draft. Writes go to your draft, not the live site.
  4. Validate. A dry run of the save checks, committing nothing.
  5. Save. Commits the draft and starts a build.
  6. Build. It checks until the build finishes, or reports the error to fix.
  7. Deploy a preview. It deploys to a preview URL and shares the link.
  8. You go live, in Portal Studio, yourself. The assistant can’t publish to your live site.
Edits are made against a specific version of the portal. If the portal changed in the meantime, for example because you edited it in Portal Studio, the write is refused rather than overwriting your work.

Reading the full content of an activity

By default, your assistant sees an activity’s overview, such as its subject, summary, and message count, but not the messages themselves. To see what was actually written, ask for the messages explicitly. For example:
  • “Show me every message in the email thread with Acme from last Tuesday.”
  • “Get the full Email Details for this activity, including the sender, time, and body of each message.”
  • “What did the customer say in the latest messages on this support ticket?”
This works for emails, Slack and Teams conversations, support tickets, Salesforce Chatter, notes, and survey responses. For each message in the thread, your assistant gets:
  • Subject, sender, recipients, and time sent
  • The message body, and the names of any attachments
  • A message ID from the source system, for example the email’s Message-ID
Messages come back oldest first. For long threads, your assistant fetches them a page at a time. If you need the whole thread, ask it to keep going until it has every message.
Telling duplicates from replies: two messages with the same message ID are copies of the same message. A reply always has its own ID. If a message has no ID (some sources don’t provide one), treat it as a separate message.
What you get is the text Statisfy captured when the activity was synced:
  • For email, this is the plain-text version, with quoted earlier replies and signatures removed. Attachments are listed by name only; their contents aren’t included.
  • bcc recipients are never returned.
  • You only see activities that your Statisfy user can see in the app.

Transport

Statisfy serves MCP over the Streamable HTTP transport from the MCP specification, on POST https://api.statisfy.com/mcp. Claude and other spec-compliant MCP clients use it automatically when you connect. The older HTTP+SSE transport isn’t supported: GET /mcp returns 405 Method Not Allowed.