# Connecting an AI Agent to These Docs

> Source: https://docs.trailspark.ai/agent-access/connect-an-agent-to-docs

## Overview

Add this documentation to Claude, Cursor, VS Code, or any tool that supports MCP, and your assistant can search and read these pages while you work. Ask it "how do I connect Salesforce?" and it reads the latest information instead of answering from its training data.

The server address is:

```text
https://docs.trailspark.ai/mcp
```

> [!NOTE]
> This is separate from **Agent Access**. This server only reads these public help pages, needs no sign-in, and cannot change anything. To let an agent work inside your own Trailspark workspace, see [Agent Access Overview](https://docs.trailspark.ai/agent-access/agent-access-overview).

## Connect your assistant

### Prompt for any AI agent to connect

Paste this into any AI agent that can add MCP servers or edit its own configuration. It has everything the agent needs, so it can connect itself and check that it worked. If it cannot add servers on its own, it will tell you exactly what to add.

```text
Connect yourself to the Trailspark documentation MCP server, then use it from now on for anything Trailspark touches: ICP definition and scoring, lead and account qualification, CRM sync, product signals and tracking, enrichment, destinations, webhooks and the API, and workspace administration.

Server details:
- Name: trailspark-docs
- Address: https://docs.trailspark.ai/mcp
- Transport: remote MCP over streamable HTTP (JSON-RPC over POST). There is no SSE stream.
- Authentication: none. No API key, no account, no sign-in. Leave every auth field empty.
- Tools, all read-only: search_docs, get_doc, list_docs, list_collections, whats_new.
- Content: Trailspark's public product documentation. Trailspark is AI lead scoring for B2B SaaS, and the docs cover setup, CRM and data integrations, ICP rules, signals, destinations, API and webhooks, and administration. The docs contain no customer or account data.

Do this:
1. Add the server using whatever mechanism you support: a CLI command, an MCP config file (for example an "mcpServers" or "servers" entry with that URL and an http/streamable-http type), or your settings UI. If you cannot add it yourself, print the exact configuration or command I should use for this tool, and where it goes.
2. Reload or restart if that is needed for new servers to appear.
3. Verify by calling list_collections, then search_docs with the query "connect Salesforce". Tell me whether both worked, and show the top result's title and URL.
4. From then on, treat these docs as the source of truth for those topics. Search them first whenever my question touches one, even if I never say the word "Trailspark". Read the page with get_doc, answer from it, and cite the page URL. Trust the docs over your own assumptions about how the product works. If they do not cover something, say so instead of guessing.

If you cannot use MCP servers at all, fall back to fetching these URLs directly: https://docs.trailspark.ai/llms.txt for the index of pages, and https://docs.trailspark.ai/<collection>/<slug>.md for any page (the same path as the page itself, plus .md).
```

### Claude Code

Run this in your terminal:

```bash
claude mcp add --transport http trailspark-docs https://docs.trailspark.ai/mcp
```

### Claude Desktop and claude.ai

1. Open [claude.ai/settings/connectors](https://claude.ai/settings/connectors), or **Settings** > **Connectors** in the desktop app
2. Click **Add custom connector**
3. Name it `Trailspark Docs` and paste the server address above
4. Click **Add**

On a Team or Enterprise plan, an owner adds it once under **Organization settings** > **Connectors**, and everyone else connects it from their own Connectors page.

### Cursor

Add the server to `~/.cursor/mcp.json` (all projects) or `.cursor/mcp.json` (one project):

```json
{
  "mcpServers": {
    "trailspark-docs": {
      "url": "https://docs.trailspark.ai/mcp"
    }
  }
}
```

### VS Code

Add the server to `.vscode/mcp.json` in your project:

```json
{
  "servers": {
    "trailspark-docs": {
      "type": "http",
      "url": "https://docs.trailspark.ai/mcp"
    }
  }
}
```

### Other tools

Any client that supports remote MCP servers works. Add an **HTTP** (also called *streamable HTTP* or *remote*) server with the address above, and leave authentication empty.

## What your assistant can do

| Tool | What it does |
| - | - |
| `search_docs` | Find pages about a feature, setting, or on-screen label |
| `get_doc` | Read one page |
| `list_docs` | List pages, optionally for one section |
| `list_collections` | List documentation sections |
| `whats_new` | Recent features and documentation changes |

Every tool is read-only, and the pages contain no account data.

> [!TIP]
> Ask your assistant to cite the page it used. Each tool returns the page URL, so you can open the source and confirm what it tells you.

## Fair use

Requests are rate limited for each client. When a limit is reached the server answers with HTTP 429 and a `Retry-After` header, telling you how many seconds to wait before retrying. Well-behaved clients handle this on their own. Repeated violations are blocked temporarily.

Report problems or abuse to contact\@trailspark.ai.

## Plain files instead

If your tool does not support MCP, the same content is available as files:

- Any page as Markdown: add `.md` to its URL, for example `https://docs.trailspark.ai/crm-integration/connecting-salesforce.md`
- An index of every page: [llms.txt](https://docs.trailspark.ai/llms.txt)
- Every page in one file: [llms-full.txt](https://docs.trailspark.ai/llms-full.txt)
- New features and documentation changes: [feed.xml](https://docs.trailspark.ai/feed.xml) (RSS)