The secure link between your AI and your Text data.

Text's MCP server is a secure connection that lets AI assistants like Claude and ChatGPT work directly with your Text data: tickets, chat transcripts, tags, and metadata. It uses the Model Context Protocol (MCP), an open standard introduced by Anthropic for connecting AI assistants to external tools and data. With Text's MCP server, your team can ask Claude or ChatGPT questions in natural language and get answers grounded in your actual support history. No copy-pasting transcripts, no exporting data, no custom integration work.

Claude MCP integration with Text takes a few minutes to set up. In Claude (Desktop, Code, or Web), add a new integration pointing at https://mcp.text.com/, authenticate via OAuth, and Claude immediately gets scoped access to your tickets, chat archives, and transcripts. From there, you can ask things like "Find tickets from this week that mention billing issues" or "Summarize the top complaints in chats tagged 'churn risk'". Claude will use the appropriate Text tool under the hood (find-tickets, get-chat-transcript, and others) to answer with real data from your account.

Yes. The ChatGPT MCP integration with Text uses the same endpoint (https://mcp.text.com/) and the same OAuth setup as Claude. Once connected, ChatGPT can search across your tickets and chats and pull back relevant context. Particularly useful for ChatGPT's Deep Research mode, which uses Text's search and fetch tools to pull data into longer analytical responses. Setup follows OpenAI's standard MCP integration flow and works against the same permissions model as the rest of Text.

An AI support agent doesn't have to mean a chatbot talking to customers. For most support teams, the bigger win is giving your internal team an AI copilot that knows your support history. By connecting Claude or ChatGPT to Text via MCP, your AI support agent can search across thousands of past conversations, pull transcripts on demand, summarize patterns, draft replies grounded in real cases, and answer questions like "how did we resolve this last time?" in seconds. It's the difference between an AI that guesses based on training data and an AI that actually knows what your team did yesterday.

The MCP integration is primarily designed to give your internal team an AI customer support agent: a copilot that helps human agents resolve tickets faster by accessing your support history through Claude or ChatGPT. For customer-facing AI that talks directly to your customers and resolves tickets end-to-end without a human in the loop, that's a different product. Text's AI Agent integrates natively with your help desk and handles whole ticket categories autonomously. Many teams use both: AI Agent for direct customer deflection, and the MCP server to give the human team an AI copilot.

Text's MCP server exposes a curated set of tools that AI assistants can call to access your data:

• Ticket operations: find-tickets (by ID, agent, or tag), list-tickets (with filtering and pagination), list-ticket-tags
• Chat operations: list-archived-chats, get-chat-transcript, list-chat-tags
• Deep Research (ChatGPT only): search and fetch for pulling relevant data into longer analytical responses

You don't need to remember the tool names. When you ask the AI assistant something in natural language, it picks the right tool automatically.

Yes. Security is the core design constraint. Three things to know:

• Permission-aware access: every MCP request respects the authenticated user's permissions in Text. The AI assistant can only see what the connected user can see.
• Strict data isolation: users can only view data they're explicitly authorized to access. No cross-account leakage.
• OAuth-based authentication (recommended): browser-based, secure, no copying tokens between systems. Bearer tokens are also supported for advanced setups.

The MCP server doesn't give AI assistants special powers. It gives them scoped access through the same permission system the rest of Text uses.

The setup takes about five minutes:

1. Open your AI assistant's integrations panel (for example, Claude Web → Settings → Integrations → Add integration).
2. Add a new MCP integration with the endpoint https://mcp.text.com/ (Claude Web uses https://mcp.text.com/mcp).
3. Authenticate via OAuth. A browser window will open to log into Text.
4. Test the connection by asking a simple question like "List my recent tickets."

Text's MCP server works with Claude Desktop, Claude Code, Claude Web, and ChatGPT today. Cursor, Windsurf, and VS Code support is coming soon.

Three different things, often confused:

• MCP server connects your AI assistant (Claude, ChatGPT) to Text data, so the AI has read access to tickets and chats. Best for: internal team copilots, ad-hoc analysis, research.
• AI Agent is Text's customer-facing AI that resolves entire ticket categories end-to-end without a human. Best for: deflecting common requests directly.
• Workflows is rules-based automation that runs deterministically on triggers and actions. Best for: predictable routing, escalation, follow-ups.

Most teams use all three: Workflows for the predictable stuff, AI Agent for direct customer deflection, MCP server for the team's own AI copilot.

Real examples teams are running today:

• Pattern detection: "Analyze last month's chats and surface the top 5 recurring pain points."
• Ticket research: "How did we handle this issue the last few times? Pull the relevant transcripts."
• Daily standups: "Show me high-priority tickets from yesterday that aren't yet resolved."
• Quality assurance: "Pull 10 random chats tagged 'refund' and assess agent tone."
• Customer-specific lookup: "Get the full conversation history for customer@example.com."

Each of these would be a multi-step task without MCP: exporting data, opening transcripts, copying into a prompt. With Text's MCP server, it's a single natural-language question.