AI agents and coding assistants

Everything on this site is published for machines as well as people. Point your coding agent at it once and it can look up the exact request shapes, error codes and limits while it writes your integration.

Connect the docs MCP server

The site hosts a Model Context Protocol server at https://docs.tawked.com/mcp. It gives an agent three tools: search the docs, read any page or section, and report a problem back to us. No key, no account.

claude mcp add --transport http tawked https://docs.tawked.com/mcp

Add to ~/.codex/config.toml:

[mcp_servers.tawked]
type = "http"
url = "https://docs.tawked.com/mcp"

Add to .cursor/mcp.json in your project, or the global mcp.json:

{
  "mcpServers": {
    "tawked": { "url": "https://docs.tawked.com/mcp" }
  }
}

Add to .vscode/mcp.json:

{
  "servers": {
    "tawked": { "type": "http", "url": "https://docs.tawked.com/mcp" }
  }
}

Add to mcp_config.json:

{
  "mcpServers": {
    "tawked": { "serverUrl": "https://docs.tawked.com/mcp" }
  }
}

Add to opencode.json:

{
  "mcp": {
    "tawked": { "type": "remote", "url": "https://docs.tawked.com/mcp", "enabled": true }
  }
}

Settings, then Connectors, Add custom connector, URL https://docs.tawked.com/mcp.

The same options sit behind the menu at the top of every page: copy the page as Markdown, open it in ChatGPT or Claude, add the MCP server to Cursor or VS Code in one click, or download the OpenAPI document from any reference page.

Install the skill

Agents that support skills can load a generated skill.md that describes what Tawked does and how the calls fit together:

npx skills add https://docs.tawked.com

The file is maintained with these pages and tested against the same source as the API reference, so it never goes stale. Read it directly at /skill.md.

Read pages as Markdown

Every page is available as clean Markdown. Append .md to its URL, or ask for it with an Accept header:

curl -L -H "Accept: text/markdown" https://docs.tawked.com/quickstart

Two index files cover the whole site: llms.txt lists every page with a description, and llms-full.txt is the whole site in one file for a context window.

The OpenAPI document

The API reference on this site is generated from https://tawked.com/openapi.json, an OpenAPI 3.1 document with every endpoint, parameter, response example and error code. Feed it to a code generator, a request client, or an agent that reads specs directly. Tawked also publishes an RFC 9727 API catalog at the well-known path.

A prompt that works

Paste this into your agent along with the MCP server or the llms-full.txt file:

Integrate Tawked OTP verification into this project.
Use the Tawked docs (MCP server or llms-full.txt) for every request shape and error code; do not guess.
Rules from the docs: authenticate with a Bearer key from an environment variable, never in code.
Call POST /v1/verify/start with `to`, `lang` and a `reference`, and send an Idempotency-Key per attempt.
Call POST /v1/verify/check with the id and the code the user typed; branch on `verified`, then on `status`.
Match errors on the `error` string, never on the HTTP status alone.
Use a tk_test_ key until the application is approved.

Agents connected to the MCP server read the docs only. They cannot send messages or spend credits; that still needs your API key in your own code.

Get started

Verify

WhatsApp

For agents

AI agents and coding assistants

Connect the docs MCP server

Install the skill

Read pages as Markdown

The OpenAPI document

A prompt that works

​Connect the docs MCP server

​Install the skill

​Read pages as Markdown

​The OpenAPI document

​A prompt that works

Connect the docs MCP server

Install the skill

Read pages as Markdown

The OpenAPI document

A prompt that works