MCP connector

Ask ChatGPT and Claude about your own company

Connect DoclyChat once and your AI assistant answers from your documents, figures and shop, with the source for every answer. No code, about three minutes.

Your AI assistant asks DoclyChat for you

You keep working where you are. When a question is about your company, the assistant gets the answer from DoclyChat.

You ask

In ChatGPT, Claude or another AI tool, the way you always do.

Your AI assistant

Sees the question is about your company and hands it to DoclyChat.

DoclyChat

Finds the answer in your knowledge base, only in what your account may see.

Answer with source

Comes back with the document and page, so you can check it.

MCP (Model Context Protocol) is the open standard AI assistants use to connect to other apps. You don't need to know more than that: you paste one link and you are done.

Everything DoclyChat knows, inside your AI assistant

Just ask in your own words. The assistant picks the right DoclyChat tool by itself.

Your documents

Contracts, manuals, policies, price lists: answers with the exact passage and page.

  • “What do our terms say about cancellation?”
  • “Find every passage about warranty in our documents”

Your figures

Annual statements and line items, with the Financial Data add-on.

  • “Compare our revenue for 2023 and 2024”
  • “What were the largest expense items last year?”

Your shop

Live stock and prices from your Shopify or WooCommerce store, and product comparisons.

  • “Is the blue hoodie in size L in stock right now?”
  • “Compare our two best-selling backpacks”

Your chatbot's conversations

Read what customers and staff asked your DoclyChat chat.

  • “Summarise yesterday's chatbot conversations”
  • “Which questions did customers ask most this week?”

Your offers and records

Imported product and offer lists, searched by meaning instead of exact words.

  • “Which of our offers fit a credit card without an annual fee?”
  • “Find products similar to the Alpine 40 backpack”

Your account

Credits left, plan, and whether new documents have finished processing.

  • “How many chat credits do we have left this month?”
  • “Has the new price list finished processing?”

Connected in four steps

One link from DoclyChat, pasted into your AI tool. That is all it takes.

  1. 1

    Copy your connector link from DoclyChat

    Sign in and open Settings → MCP server. Copy the Connector URL for ChatGPT & Claude.ai. Your key is already built in, so this one link is all you need.

  2. 2

    Pick your AI tool

    ChatGPT and Claude only need the link. Developer tools take a short config file, which we prepare for you in step 3.

  3. 3

    Paste it in

    Paste the key from Settings → API key and every snippet below fills itself in, ready to copy. For ChatGPT and Claude the connector link from step 1 already has it. The key stays in your browser; this page never sends it anywhere.

    ChatGPT

    ChatGPT Plus, Pro, Business, Enterprise and Edu on chatgpt.com. Custom connectors run through ChatGPT's developer mode.

    1. Open Settings → Apps & Connectors → Advanced settings and turn on Developer mode. On Business/Enterprise a workspace admin may have to allow custom connectors first.
    2. Back in Apps & Connectors, click Create.
    3. Name it DoclyChat, paste the MCP Server URL shown here, and set Authentication to No authentication – the key at the end of the URL is the login.
    4. Confirm you trust the connector and click Create. ChatGPT lists the DoclyChat tools it found.
    5. In a new chat, open the + menu, pick Developer mode, switch DoclyChat on and ask, e.g. "Ask DoclyChat what our return policy is."
    MCP Server URL
    https://doclychat.com/mcp/<your-api-key>

    Claude.ai / Claude Desktop

    Claude Pro, Max, Team and Enterprise. Added once, the connector works on claude.ai, in Claude Desktop and in the mobile apps.

    1. Open Settings → Connectors. On Team/Enterprise an owner adds it under Admin settings → Connectors first.
    2. Click Add custom connector.
    3. Name it DoclyChat and paste the Remote MCP server URL shown here. Leave the OAuth fields under Advanced settings empty.
    4. Click Add. In a chat, open the tools menu and make sure DoclyChat is switched on.
    Remote MCP server URL
    https://doclychat.com/mcp/<your-api-key>

    Claude Code

    Anthropic's coding agent for the terminal. Add --scope user to make DoclyChat available in every project.

    1. Copy the command and run it in your terminal.
    2. Run claude mcp list – doclychat should show as connected.
    3. Inside Claude Code, type /mcp to see the DoclyChat tools, then just ask.
    Terminal
    claude mcp add --transport http doclychat https://doclychat.com/mcp \
      --header "X-Api-Key: <your-api-key>"

    Cursor

    Global in ~/.cursor/mcp.json, or per project in .cursor/mcp.json.

    1. Open Cursor Settings → MCP & Integrations → New MCP server – this opens mcp.json.
    2. Paste the JSON (merge it into mcpServers if the file already has entries) and save.
    3. DoclyChat turns green in the MCP list and its tools are available to the agent in chat.
    ~/.cursor/mcp.json
    {
      "mcpServers": {
        "doclychat": {
          "url": "https://doclychat.com/mcp",
          "headers": { "X-Api-Key": "<your-api-key>" }
        }
      }
    }

    VS Code (Copilot)

    GitHub Copilot Chat in agent mode.

    1. Create .vscode/mcp.json in your workspace, or run MCP: Add Server from the Command Palette for all workspaces.
    2. Paste the JSON and click the Start link that appears above the server.
    3. Open Copilot Chat, switch to Agent mode and enable the DoclyChat tools in the tools picker.
    .vscode/mcp.json
    {
      "servers": {
        "doclychat": {
          "type": "http",
          "url": "https://doclychat.com/mcp",
          "headers": { "X-Api-Key": "<your-api-key>" }
        }
      }
    }

    Windsurf

    Cascade reads ~/.codeium/windsurf/mcp_config.json.

    1. In Cascade, open the MCPs menu → Manage MCPs → View raw config.
    2. Paste the JSON and save.
    3. Click Refresh – the DoclyChat tools appear in Cascade.
    ~/.codeium/windsurf/mcp_config.json
    {
      "mcpServers": {
        "doclychat": {
          "serverUrl": "https://doclychat.com/mcp",
          "headers": { "X-Api-Key": "<your-api-key>" }
        }
      }
    }

    Gemini CLI

    Google's Gemini CLI. Use .gemini/settings.json inside a project to limit it to that project.

    1. Open (or create) ~/.gemini/settings.json.
    2. Paste the JSON – merge it into mcpServers if that already exists.
    3. Start gemini and run /mcp to check DoclyChat is connected.
    ~/.gemini/settings.json
    {
      "mcpServers": {
        "doclychat": {
          "httpUrl": "https://doclychat.com/mcp",
          "headers": { "X-Api-Key": "<your-api-key>" }
        }
      }
    }

    OpenAI Codex

    OpenAI's Codex CLI and IDE extension share ~/.codex/config.toml.

    1. Open ~/.codex/config.toml.
    2. Add the block shown here. It uses the URL with the key at the end, so no extra auth settings are needed. On older Codex versions also add experimental_use_rmcp_client = true at the top of the file.
    3. Restart Codex and run /mcp to see the DoclyChat tools.
    ~/.codex/config.toml
    [mcp_servers.doclychat]
    url = "https://doclychat.com/mcp/<your-api-key>"

    OpenAI / Claude API

    Building your own agent? Both APIs call remote MCP servers directly – no client code to write.

    1. OpenAI Responses API: add the first block to the tools array. The OpenAI Agents SDK takes the same settings in its hosted MCP tool.
    2. Claude Messages API: add the second block to mcp_servers using Anthropic's MCP connector (see their docs for the current beta header). The key is sent as a bearer token, which DoclyChat accepts.
    OpenAI Responses API -- tools[]
    {
      "type": "mcp",
      "server_label": "doclychat",
      "server_url": "https://doclychat.com/mcp",
      "headers": { "X-Api-Key": "<your-api-key>" },
      "require_approval": "never"
    }
    Claude Messages API -- mcp_servers[]
    {
      "type": "url",
      "name": "doclychat",
      "url": "https://doclychat.com/mcp",
      "authorization_token": "<your-api-key>"
    }

    n8n, Le Chat & others

    n8n, Make, Zapier, Mistral Le Chat, Microsoft Copilot Studio, LibreChat, Open WebUI – anything that supports remote MCP over Streamable HTTP.

    1. If the tool lets you set a header, API key or bearer token, use https://doclychat.com/mcp with X-Api-Key (or Authorization: Bearer) set to your key.
    2. If it only asks for a server URL, use the URL-only form with the key at the end and choose no authentication.
    3. Pick Streamable HTTP as the transport if asked. A tool that only runs local (stdio) servers can bridge with npx mcp-remote https://doclychat.com/mcp/<key>.
    Header auth
    URL:     https://doclychat.com/mcp
    Header:  X-Api-Key: <your-api-key>
             (or Authorization: Bearer <your-api-key>)
    URL only
    https://doclychat.com/mcp/<your-api-key>
  4. 4

    Ask your first question

    Open a new chat and try one of these. Click a question to copy it.

    How you know it works
    • Your AI tool shows that it is calling DoclyChat, usually as a small tool note above the answer.
    • The answer names a document and page from your own files.
    • Nothing happens? In ChatGPT and Claude, switch DoclyChat on in that chat's tools or + menu first.

We help you get connected

Most people are done in a few minutes. If something gets stuck, this is where to look.

Ask us directly

Write to [email protected] or call +49 831 52 71 50 82, tell us which AI tool you use, and we will help you get connected. You can also ask the chat on this page.

Not working?

  • 401 / unauthorized – the key is wrong or was regenerated. Copy it again from Settings → API.
  • No Create button in ChatGPT – turn on Developer mode first, or ask your workspace admin to allow custom connectors.
  • Connected but no tools – restart the app or reload its MCP servers; most tools read their config only at startup.
  • Answers say there's no data – the key belongs to a different DoclyChat account, or the documents aren't uploaded yet.

Two ways to send your key

  • Header (use it wherever the tool allows): X-Api-Key: <key> or Authorization: Bearer <key> to https://doclychat.com/mcp.
  • Secret URL for tools that only take a URL (ChatGPT, Claude.ai): https://doclychat.com/mcp/<key>. Treat it like a password.
  • Shared it by mistake? Regenerate the key in Settings → API – every old URL and header stops working at once.

Test the connection yourself

The official MCP Inspector lists DoclyChat's tools and lets you call them by hand. Run it, choose Streamable HTTP, paste https://doclychat.com/mcp/<your key> and click Connect.

Terminal
npx @modelcontextprotocol/inspector
Do I need to be technical?

No. In ChatGPT and Claude you paste one link and click Add. Config files are only for developer tools like Cursor or VS Code, and the setup guide above prepares them for you.

Which ChatGPT and Claude plans work?

ChatGPT Plus, Pro, Business, Enterprise and Edu, through developer mode, and Claude Pro, Max, Team and Enterprise. On team plans an admin may have to allow custom connectors first.

Is my data safe?

Your files stay in DoclyChat. For each question your AI tool only receives the answer and passages DoclyChat sends back, handled under that tool's own terms. Treat the connector link like a password: anyone who has it can ask your DoclyChat. If it ever leaks, click Regenerate next to the API key in Settings and the old link stops working.

Does it cost extra?

No. Questions answered through MCP use your chat credits, just like questions in the DoclyChat chat.

Can I use several AI tools at once?

Yes. The same link or key works in all of them at the same time, for example ChatGPT on your laptop and Claude on your phone.

How is this different from the chat on my website?

The website chat is for your customers. MCP is for you and your team: it brings DoclyChat into the AI assistant you already work in. Both answer from the same documents.

Every MCP tool, at a glance

All tools are called the same way – an X-Api-Key header carrying your tenant's key – and every one that hits a snag returns a plain { error } instead of a cryptic status code. Click any tool below for its parameters, things you can ask, and a real example call and response.

ask_chatbot Ask the chatbot a question exactly like the chat widget would – includes guaranteed-JSON slash commands such as /toplist and /compare.
Parameters
  • question (string, required) – plain-language text, or one of the slash commands below for a guaranteed JSON shape instead of free text.
Try asking
  • "What's our refund policy for damaged items?"
  • "How do I reset my password?"
  • "/toplist our best-selling products"
  • "/compare Starter Plan vs Pro Plan"
Example call
{ "question": "What is our return policy?" }
Example response
{
  "answer": "Items can be returned within 30 days...",
  "sources": [
    { "document": "Returns-Policy.pdf", "page": 1, "snippet": "..." }
  ]
}
search_knowledge_base Raw passage search over the tenant's knowledge base – no composed answer, just the matching source text.
Parameters
  • query (string, required) – natural language or keywords.
  • limit (integer, optional, default 10, max 50).
Try asking
  • "Find every mention of 'late payment fee' in our contracts"
  • "Search for 'warranty period' across all our documents"
Example call
{ "query": "late payment fee", "limit": 5 }
Example response
{
  "results": [
    { "document": "Terms.pdf", "page": 4, "text": "...", "score": 0.87, "sourceUrl": null }
  ]
}
get_financial_statements Lists ingested financial statements and their line items (Financial Data add-on).
Parameters
  • take (integer, optional, default 20, max 100) – how many statements to return, most recent first.
Try asking
  • "Show me our last 5 financial statements"
  • "What line items were in our 2024 statement?"
Example call
{ "take": 5 }
Example response
[
  {
    "id": "3fa85f64-...",
    "document": "Q4-2024.pdf",
    "fiscalYear": 2024,
    "status": "Completed",
    "lineItemCount": 42,
    "lineItems": [
      { "account": "Revenue", "category": "Income", "amount": 128500.00 }
    ]
  }
]
get_usage_status Current AI credit standing for the billing period.
Parameters
  • No parameters.
Try asking
  • "How many AI credits have we used this month?"
  • "Are we close to our credit limit?"
Example call
{ }
Example response
{
  "monthlyLimit": 50000,
  "creditsUsedThisPeriod": 17,
  "bonusCredits": 0,
  "isNearLimit": false,
  "isExceeded": false
}
get_subscription_status Plan, credit usage, and auto-recharge settings – read-only.
Parameters
  • No parameters.
Try asking
  • "What plan are we currently on?"
  • "Is auto-recharge turned on for our account?"
Example call
{ }
Example response
{
  "plan": { "key": "pro", "name": "Pro", "monthlyCredits": 50000 },
  "usage": { "monthlyLimit": 50000, "creditsUsedThisPeriod": 17 },
  "autoRecharge": { "enabled": false, "thresholdCredits": 1000 }
}
list_conversations / get_conversation List every user's chat conversations, or read one conversation's full message history by id.
Parameters
  • list_conversations: includeArchived (boolean, optional, default false), take (integer, optional, default 20, max 100).
  • get_conversation: conversationId (string, required) – from list_conversations.
Try asking
  • "What questions did people ask the chatbot this week?"
  • "Show me the full transcript of conversation abc-123"
Example call
{ "take": 10 }
Example response
{
  "conversations": [
    { "id": "abc-123", "title": "Return policy question", "updatedAtUtc": "...", "isPinned": false }
  ]
}
get_ingestion_status Sync status of every connected content source – Drive, SharePoint, Shopify, and more.
Parameters
  • No parameters.
Try asking
  • "Is our Google Drive sync working?"
  • "When did our Shopify connector last sync, and did it fail?"
Example call
{ }
Example response
{
  "sources": [
    { "connectorType": "GoogleDrive", "connectorName": "Client Docs", "enabled": true,
      "lastSyncedAtUtc": "...", "lastSyncStatus": "Succeeded", "lastSyncError": null }
  ]
}
list_enabled_agents Which optional add-ons (Financial Data, HR, Heizungsbauer, ...) are enabled for this tenant.
Parameters
  • No parameters.
Try asking
  • "Which add-ons do we currently have turned on?"
Example call
{ }
Example response
{
  "agents": [
    { "key": "financial-data", "name": "Financial Data", "description": "..." }
  ]
}
check_shop_live_data Live price/stock lookup straight from a connected Shopify or WooCommerce store.
Parameters
  • query (string, required) – a product name, SKU, or stock/price question.
Try asking
  • "How many units of SKU-1042 do we have in stock right now?"
  • "What's the current live price of the Blue Widget?"
Example call
{ "query": "SKU-1042 stock" }
Example response
{
  "results": [
    { "shop": "My Shopify Store", "connectorType": "Shopify", "result": { "sku": "SKU-1042", "stock": 14 } }
  ]
}
estimate_project_cost Statistical HVAC job cost estimate from historical invoices (Heizungsbauer add-on).
Parameters
  • description (string, required) – a concise description of the job.
  • maxComparableItems (integer, optional, default 200).
Try asking
  • "Estimate the cost of replacing a gas boiler in a single-family home"
Example call
{ "description": "Austausch einer Gastherme in einem Einfamilienhaus" }
Example response
{
  "hasEnoughData": true,
  "suggestedTotalEstimate": 3200.00,
  "groups": [
    { "matchedWork": "Gastherme Austausch", "count": 14, "averagePrice": 3200, "minPrice": 2500, "maxPrice": 4100 }
  ]
}
query_hr_information Search HR-tagged documents only – handbooks, policies, contracts (HR add-on).
Parameters
  • query (string, required).
  • limit (integer, optional, default 10, max 50).
Try asking
  • "What's our parental leave policy?"
  • "How many vacation days do new hires get?"
Example call
{ "query": "parental leave policy" }
Example response
{
  "results": [
    { "document": "Employee-Handbook.pdf", "page": 12, "text": "...", "score": 0.91 }
  ]
}
create_sub_tenant Provisions a brand-new, fully independent tenant – see "Build a platform on DoclyChat" below.
Parameters
  • tenantName (string, required) – display name for the new tenant.
Try asking
  • Not a conversational tool – called programmatically by a partner platform whenever it needs to provision a new client tenant. See "Build a platform on DoclyChat" below.
Example call
{ "tenantName": "Acme Corp" }
Example response
{
  "tenantId": "...",
  "tenantName": "Acme Corp",
  "apiKey": "... (shown once, store it now)"
}

Build a platform on DoclyChat

Running your own product on top of DoclyChat? create_sub_tenant lets your own API key provision an isolated tenant – its own knowledge base, its own credits – for each of your downstream clients, no DoclyChat admin involved.

1

1. Provision

Call create_sub_tenant with your own API key and a name for the client. You get back a tenantId and a brand-new apiKey – store it immediately, it's shown once.

2

2. Ingest

Push each of that client's documents into their new tenant with POST /api/knowledge/documents/ingest, using the client's own apiKey.

3

3. Query

Ask questions or search that client's data over /mcp or REST, using the same apiKey – fully isolated from every other tenant, including yours.

create_sub_tenant
// request
{ "tenantName": "Acme Corp" }

// response -- store apiKey now, it is never shown again
{ "tenantId": "...", "tenantName": "Acme Corp", "apiKey": "..." }

Each sub-tenant is headless (no email/password login) – the apiKey returned once is the only way back in, so store it immediately.

Your AI assistant, now with your company's knowledge

Start free, copy your connector link from Settings → MCP server and paste it into ChatGPT or Claude.

English German
English German
An unhandled error has occurred. Reload 🗙

Rejoining the server...

Rejoin failed. Retrying in 0s (attempt 1)

Failed to rejoin. Please retry or reload the page.

Your session was paused by the server.

Failed to resume the session. Please retry or reload the page.

Your session has expired. Reloading…