> ## Documentation Index
> Fetch the complete documentation index at: https://apidocs.writesonic.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents MCP Server

> Connect Writesonic Agents to Claude, ChatGPT, and other MCP-compatible assistants. Manage tasks, playbooks, grids, knowledge, and content intelligence from chat.

## Overview

The Writesonic **Agents MCP** server lets an AI assistant work with your [Agents](/agents/overview) workspace directly from a conversation. Ask it to list your tasks, create one, inspect a playbook or grid, search your knowledge sources, or pull a content-intelligence report, and it calls the right tool for you.

**Server URL:**

```
https://mcp.agent.writesonic.com/mcp
```

<Info>
  **Agents is in beta.** Access is currently limited to selected enterprise and agency accounts. To request it, talk to your account manager or [book a demo](https://writesonic.com/?demo=open) with our sales team.
</Info>

<Note>
  This is a different server from the [GEO MCP](/mcp/mcp-server) at `mcp.writesonic.com`. The GEO server answers questions about your AI-search **visibility data**; the Agents server operates your Agents **workspace**: projects, tasks, playbooks, grids, and knowledge.
</Note>

## What You Can Do

* **Manage tasks**: create tasks, query and filter them, transition status, assign owners, and comment
* **Track agent work**: pull agent-run history and step-by-step progress for a task or agent
* **Inspect playbooks**: browse a project's playbooks, versions, runs, and the building-block components
* **Read grids**: list grids and read their columns, rows, cells, and cell run detail
* **Search knowledge**: run semantic search across your project's knowledge sources, and read the Brand Kit
* **Pull content intelligence**: get the page dashboard and per-page SEO, citation, funnel, traffic, and SERP reports
* **See your setup**: resolve your identity and projects, list integrations, users, labels, skills, and agents
* **Handle attachments**: get short-lived download URLs for task and comment attachments

## Connection Methods

| Platform           | How to Connect                                                                                        |
| ------------------ | ----------------------------------------------------------------------------------------------------- |
| **Claude.ai**      | Settings › Connectors › Add custom connector, `https://mcp.agent.writesonic.com/mcp`                  |
| **Claude Desktop** | Add to `claude_desktop_config.json` (see [Setup Guide](/agents-mcp/setup))                            |
| **Claude Code**    | `claude mcp add writesonic-agents --transport http https://mcp.agent.writesonic.com/mcp`              |
| **ChatGPT**        | Settings › Connectors › enable Developer Mode › Create (see [Setup Guide](/agents-mcp/setup#chatgpt)) |

Team and Enterprise admins can roll the connector out org-wide, see [Organization Setup](/agents-mcp/setup#organization-setup).

## How It Works

1. **Connect**: add the Agents MCP server URL to your assistant
2. **Authenticate**: sign in with your Writesonic account via OAuth when prompted
3. **Work**: ask in natural language; the assistant calls the tools

### Example Conversation

> **You:** What tasks are open in my project, and who's on them?
>
> **Assistant:** Let me check your Agents workspace.
>
> *Assistant calls `whoami`, then `query_tasks` with `view: "list"` scoped to your project.*

## Scoping: projects come first

Almost every tool needs a `project_id`. Start each session by having the assistant call `whoami`, it returns your identity, your teams, and the projects in your active workspace. The assistant uses a project's `id` from that response for every following call. If more than one project could match, the assistant should ask rather than guess.

## Requirements

* A Writesonic account with **Agents access** (beta)
* At least one project in your [Agents workspace](https://agents.writesonic.com)

## Available Tools

The server exposes 20 tools for reading and operating your Agents workspace, plus two meta-tools (`execute_batch`, `execute_plan`) for chaining calls efficiently. See the [Tools Reference](/agents-mcp/tools-reference) for full details. The core tools to know:

| Tool                                   | Description                                          |
| -------------------------------------- | ---------------------------------------------------- |
| `whoami`                               | Your identity, teams, and projects (call this first) |
| `query_tasks`                          | List or count a project's tasks                      |
| `get_task`                             | Full detail for one task                             |
| `create_task`                          | Create a task in the project                         |
| `update_task`                          | Edit fields, transition status, or assign a task     |
| `comment_on_task`                      | Comment on a task                                    |
| `get_agent_runs`                       | Agent-run history and per-run progress               |
| `list_workflows` / `get_workflow`      | Browse playbooks and their versions and runs         |
| `list_grids` / `get_grid_data`         | List grids and read their data                       |
| `search_knowledge`                     | Semantic search over knowledge sources               |
| `get_brand_kit`                        | Read the project's brand kit                         |
| `get_ci_dashboard` / `get_page_report` | Content-intelligence dashboard and page reports      |

## Next Steps

<CardGroup cols={2}>
  <Card title="Setup Guide" icon="gear" href="/agents-mcp/setup">
    Platform-specific setup instructions
  </Card>

  <Card title="Tools Reference" icon="wrench" href="/agents-mcp/tools-reference">
    Every tool, its purpose, and its REST equivalent
  </Card>

  <Card title="Use Cases & Examples" icon="lightbulb" href="/agents-mcp/examples">
    Real-world prompts and use cases
  </Card>

  <Card title="Agents REST API" icon="code" href="/api-reference/agents/authentication">
    Call the same tools over HTTP
  </Card>
</CardGroup>