Skip to content
Get Started

Claude connector

Add intoCHAT to Claude and work on your agents in plain words

intoCHAT has a remote MCP server. Add it to Claude as a custom connector, sign in to intoCHAT, and Claude can list your agents, read their conversations and leads, add Q&A pairs and test replies for you.

Create your agent free See it in the docs

Last updated:

What the Claude connector is, and what it is not

MCP (Model Context Protocol) is the standard way AI tools such as Claude connect to other services. intoCHAT runs a remote MCP server at the address below. When you add it to Claude as a custom connector and sign in, Claude gets a set of intoCHAT tools it can call while you chat with it: "Which questions did my support agent get this week?", "Show me yesterday's leads", "Add a Q&A pair about our new return window and test how the agent answers it." This is different from running your website chatbot on a Claude model. On intoCHAT's paid plans you can also choose Claude as the model that writes your agent's replies to visitors; that is a setting on the agent, and the Claude chatbot feature page covers it. The connector works the other way round: Claude, the assistant you use yourself, reaches into your intoCHAT account. intoCHAT isn't listed in Claude's connector directory. You add it yourself as a custom connector, in a few steps: paste the server address, sign in to intoCHAT, and choose what Claude may do. No API key to copy. The connector you add in Claude.ai also works in the Claude desktop app. For Claude Code, you connect with an intoCHAT API key instead. A few ways to use it: a Monday review where Claude reads last week's conversations and lists the questions the agent couldn't answer well; turning a policy change into a Q&A pair and checking the agent's new answer with a test message; or pulling the week's leads into a summary you paste into your CRM. Claude does the reading and writing; you decide what it may touch. The MCP server is included on Starter, Pro and Enterprise, like the REST API. Every tool does exactly what the matching REST endpoint does, and the connection can never do more than you can do in the account yourself.

Add intoCHAT to Claude as a custom connector

  1. Open Connectors in Claude

    In Claude.ai, open Settings, then Connectors, and click Add custom connector. Depending on your Claude plan, an organization owner may have to add it for everyone instead.

  2. Paste the server address

    Give the connector a name, for example intoCHAT, and paste https://www.intochat.ai/api/mcp as the URL. Then click Add, and Connect. A new window opens on intoCHAT.

  3. Sign in to intoCHAT

    Sign in if you aren't signed in yet; you come back to the consent screen afterwards. Check the app's name and the address it sends you back to.

  4. Choose what Claude may do

    Pick the Account if you belong to more than one, then Access (Read only, or Read and write) and Agents (All agents, or Only the agents I choose). An account on Free can't be connected.

  5. Click Allow

    You're sent back to Claude.ai and the connector shows as connected. Ask Claude, for example: "List my intoCHAT agents."

The connector URL

https://www.intochat.ai/api/mcp

Use the address with www, as shown. In Claude Code, add it with an intoCHAT API key: claude mcp add --transport http intochat https://www.intochat.ai/api/mcp --header "Authorization: Bearer ic_live_YOUR_KEY"

What Claude can do in your intoCHAT account

  • Read your agents

    list_agents and get_agent return your agents with their IDs, training status and counts, so Claude knows which agent you mean.

  • Read conversations and the Inbox

    list_conversations and get_conversation read transcripts, newest first. list_inbox_conversations lists open, pending, snoozed or solved Inbox conversations.

  • Read and update leads and contacts

    list_leads and list_contacts read what your agents collected. With write access, upsert_contact creates or updates a contact by external ID or email.

  • Teach the agent new answers

    With write access, add_text_source and add_qa_pair add knowledge and start training; list_sources and delete_source show and remove sources.

  • Test how an agent replies

    chat_with_agent sends a real message, answered like a visitor's, from the agent's knowledge, instructions and actions. Each reply uses messages from your plan.

  • Access you control

    Read only or read and write, all agents or the ones you tick, and never more than your own role allows. Revoke it any time under Settings, Account, Connected apps.

Frequently asked questions

Is intoCHAT in Claude's connector directory?

No. You add intoCHAT as a custom connector with the server address https://www.intochat.ai/api/mcp. Sign-in works through OAuth, so there is no key to paste into Claude.

Which intoCHAT plan do I need?

Starter, Pro or Enterprise. The MCP server is included with the REST API from Starter. On Free you can't connect apps, and an account that moves back to Free gets an error on every request until it is on Starter or higher again.

Which Claude plan do I need?

Custom connectors are a Claude feature, so what you can add depends on your Claude plan. On some plans an organization owner adds custom connectors for everyone. Check Settings, Connectors in Claude to see what your plan allows.

Can Claude change my agent's settings or instructions?

No. There are no tools to create agents or change their settings or instructions; do that in the intoCHAT dashboard. With write access, Claude can add text snippets and Q&A pairs, delete sources, update contacts and send test messages.

Can a teammate connect Claude too?

Yes. Anyone on the team can connect, and the connection can never do more than they can. A Viewer who allows read and write still can't add or delete sources, and someone limited to selected agents reaches only those. If their role changes, the connection changes with it on its next request.

Does using the connector cost messages?

Only chat_with_agent does: each reply counts as one message on your plan with the default model, more with a larger model, and counts toward the agent's monthly message cap if you set one. Reading conversations, leads and sources doesn't use messages.

Are there rate limits?

Each connected app can make 60 requests a minute. Connecting takes a few requests, and so does every tool call. Over the limit, requests are refused for the rest of the minute. A test reply must finish within 55 seconds.

Can I use intoCHAT in Claude Code?

Yes, with an intoCHAT API key instead of signing in. The account owner or an admin creates one under Settings, API keys, ideally with read-only scopes if Claude Code should only read. Then run the claude mcp add command shown above with your key. It adds the server for the current project; add --scope user to use it everywhere, and type /mcp in Claude Code to check that intochat is connected.

How is this different from a Claude chatbot on my website?

The connector lets you, in Claude, work with your intoCHAT account. A Claude chatbot is your intoCHAT agent answering visitors with a Claude model, which you pick per agent on a paid plan. The two are independent: you can use the connector whichever model your agents run on.

How do I disconnect Claude?

In intoCHAT, go to Settings, Account, find Claude under Connected apps and click Revoke. Access ends immediately. Connecting and disconnecting are recorded in the audit log. A connection that isn't used for 30 days expires, and you connect again.

See it answer from your own website

Paste your website address and chat with an agent built from your pages. It takes about a minute.

Create your agent free

Free plan, no credit card needed.