Use these docs with AI

Connect Cursor, Claude Code, and similar tools to the live Otter guides and API reference — so generated code matches the current contract, not training-data guesses.

Use these docs programmatically through AI assistants, code editors, and Model Context Protocol (MCP). The MCP is read-only documentation search. It does not call the Otter API or access your account.

Quick access

Start with one of these:

• Fetch llms.txt first. That manifest links the OpenAPI contract, the full guide corpus, and the MCP endpoint.
• Connect the MCP server at https://connect.tryotter.com/docs/mcp (Streamable HTTP).
• If your tool cannot use MCP, follow the static artifact URLs in llms.txt.

Pages across this site also expose the same discovery links in HTML (<link rel="alternate">) and in robots.txt.

Use our MCP server

One server covers both layers: guides (the why and how) and the OpenAPI contract (exact paths, scopes, and payloads).

Server	URL	What it covers
otter-docs	https://connect.tryotter.com/docs/mcp	Integration guides, API operations, webhook payloads, and shared schemas

Tools: search_docs, get_guide, get_openapi. After connect, the client shows the server as otter-partner-docs.

Connect with Claude Code

Run this in the project directory:

Bash

LanguageBash

claude mcp add --transport http otter-docs https://connect.tryotter.com/docs/mcp

That add is project-scoped. To use the server in every project:

Bash

LanguageBash

claude mcp add --transport http otter-docs --scope user https://connect.tryotter.com/docs/mcp

Connect with Claude Desktop

1. Open Claude Desktop.
2. Go to Settings → Connectors.
3. Add a custom connector named otter-docs with URL https://connect.tryotter.com/docs/mcp.

Or merge this into .mcp.json in the project root, or into Claude Desktop claude_desktop_config.json:

JSON

LanguageJSON

{
  "mcpServers": {
    "otter-docs": {
      "type": "http",
      "url": "https://connect.tryotter.com/docs/mcp"
    }
  }
}

Connect with Codex CLI

Bash

LanguageBash

codex mcp add otter-docs --url https://connect.tryotter.com/docs/mcp

Connect with Cursor

Save as .cursor/mcp.json in your project, or merge into ~/.cursor/mcp.json:

JSON

LanguageJSON

{
  "mcpServers": {
    "otter-docs": {
      "url": "https://connect.tryotter.com/docs/mcp"
    }
  }
}

Restart Cursor after you save.

Connect with VS Code

Save as .vscode/mcp.json:

JSON

LanguageJSON

{
  "servers": {
    "otter-docs": {
      "type": "http",
      "url": "https://connect.tryotter.com/docs/mcp"
    }
  }
}

Generic Streamable HTTP

If your client only needs the URL:

Text

LanguageText

https://connect.tryotter.com/docs/mcp

Verify

After the client connects, you should see a server named otter-partner-docs with three tools. A typical initialize payload looks like:

JSON

LanguageJSON

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "protocolVersion": "2024-11-05",
    "serverInfo": {
      "name": "otter-partner-docs",
      "version": "1.0.0"
    },
    "capabilities": {
      "tools": {}
    }
  }
}

tools/list should return search_docs, get_guide, and get_openapi.

Smoke-test from the client chat:

1. Ask: Use search_docs to find ping.
2. Expect a JSON result with results that include a ping operation or webhook and a url under https://connect.tryotter.com/docs/.
3. Ask: Call get_openapi.
4. Expect a JSON body like this (the version string comes from the live contract):

JSON

LanguageJSON

{
  "url": "https://connect.tryotter.com/docs/openapi/public-api.yaml",
  "openapiVersion": "v1"
}

5. Ask: Call get_guide with slug guides-authentication.
6. Expect the Authentication guide body and a canonical url.

Then implement against the linked guide or API reference page — not against memorized paths.

Common failures

Symptom	Likely cause	Fix
Initialize hangs	Wrong transport or a trailing path segment	Use exactly https://connect.tryotter.com/docs/mcp over Streamable HTTP, not stdio
Server connects, no tools	Client did not reload config	Restart the client after you save the file
Answers ignore the contract	Agent guessed from training data	Ask it to call search_docs, then open the returned URL before writing code

Static artifacts

These are real files on the docs host. Use the URL in the table — do not add a trailing slash.

Artifact	What it is	URL
llms.txt	Discovery manifest — start here	https://connect.tryotter.com/docs/llms.txt
llms-full.txt	All integration guides in one file	https://connect.tryotter.com/docs/llms-full.txt
agent-index.json	Stable guide URLs plus MCP metadata	https://connect.tryotter.com/docs/agent-index.json
agent-search.json	Search index over guides and OpenAPI	https://connect.tryotter.com/docs/agent-search.json
public-api.yaml	Canonical OpenAPI contract	https://connect.tryotter.com/docs/openapi/public-api.yaml
robots.txt	Crawler allow-list and artifact pointers	https://connect.tryotter.com/docs/robots.txt

Per-guide Markdown also lives at https://connect.tryotter.com/docs/agent-guides/<slug>.md.

Next

• Complete the Quickstart before building your first integration.
• Understand authentication before requesting access tokens.
• Explore the API reference for the current contract.