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

# MCP Todo Tools

> Four MCP tools any client can call to read and mutate the same todo store as the runtime

The `praisonai-mcp` host exposes four MCP tools that read and write the **same** todo store as the agent runtime (`todo_add` / `todo_list` / `todo_update`) — a todo added in Claude Desktop is visible to the runtime on the same machine, and vice versa.

```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import Agent

agent = Agent(name="Planner", instructions="Break work into small todos.")
agent.start("Plan a REST API")
```

Run `praisonai-mcp serve --transport stdio`, point Claude Desktop at it, and the assistant can list, add, complete, and delete the same todos.

```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
graph LR
    Client[📋 MCP Client<br/>Claude Desktop / Cursor] --> Host[⚙️ praisonai-mcp]
    Host --> Tools[🛠️ praisonai.todo.*]
    Tools --> Store[(💾 Shared store<br/>~/.praisonai/todos.json)]
    Runtime[🤖 Agent runtime<br/>todo_add / todo_list] --> Store

    classDef client fill:#6366F1,stroke:#7C90A0,color:#fff
    classDef heavy fill:#8B0000,stroke:#7C90A0,color:#fff
    classDef tools fill:#189AB4,stroke:#7C90A0,color:#fff
    classDef store fill:#F59E0B,stroke:#7C90A0,color:#fff
    classDef ok fill:#10B981,stroke:#7C90A0,color:#fff

    class Client client
    class Host heavy
    class Tools tools
    class Store store
    class Runtime ok
```

## Quick Start

**1. Serve the host**

```bash theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
praisonai-mcp serve --transport stdio
```

**2. Add a todo from an agent run**

```python theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
from praisonaiagents import Agent

agent = Agent(name="Planner", instructions="Break work into small todos.")
agent.start("Plan the release")
```

**3. Read it back from your client**

In Claude Desktop:

> List my todos.

The assistant calls `praisonai.todo.list` and sees the todo the runtime wrote.

***

## Tools

Four tools share the runtime's store and record schema (`id` int, `task`, `priority`, `category`, `status`).

| Tool | Arguments | Description |
| - | - | - |
| `praisonai.todo.list` | — | List every todo. Returns `"No todos found"` when the store is empty. |
| `praisonai.todo.add` | `content` (str), `priority` (str, default `"medium"`) | Append a todo. Returns `"Todo added: <id>"` where `id` is `max(existing) + 1` (starts at `1`). |
| `praisonai.todo.complete` | `todo_id` (str) | Mark the matching todo `completed`. `todo_id` is compared to the runtime's integer id as a string, so both `"1"` and `1` match. Returns `"Todo completed: <id>"`, or `"Todo not found: <id>"` if no match. |
| `praisonai.todo.delete` | `todo_id` (str) | Delete the matching todo. Returns `"Todo deleted: <id>"`, or `"Todo not found: <id>"` when nothing matched — a real error, not the silent no-op of the old code. |

> [!NOTE]
>
> The four tools were rebound to the runtime store in PraisonAI PR [#4962](https://github.com/MervinPraison/PraisonAI/pull/4962). Before that fix they wrote to `~/.praison/todo.json` in an incompatible schema (`id` as a UUID string, field name `content`) and were completely invisible to the runtime. If your workflow depended on the old path, migrate any records into `~/.praisonai/todos.json` (or your workspace's `todos.json`).

***

## Record Shape

The store is a JSON array — each todo carries an integer `id` and a `task` field.

```json theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
[
  {
    "id": 1,
    "task": "Ship the release",
    "priority": "high",
    "category": "general",
    "status": "pending",
    "created_at": "2026-09-08T15:00:00Z"
  }
]
```

Same shape the agent runtime writes — MCP-created and runtime-created todos read each other's records field-for-field.

***

## Store Location

The MCP tools resolve the same file path as `praisonaiagents.tools.todo_tools.TodoTools._get_todo_file()`.

| When | Path |
| - | - |
| A workspace is active | `<workspace>/todos.json` |
| No workspace | `~/.praisonai/todos.json` |

Never `~/.praison/todo.json` — that was the old wrong path and is no longer read or written by any surface.

***

## End-to-End Flow

```mermaid theme={"theme":{"light":"vitesse-light","dark":"vitesse-dark"}}
sequenceDiagram
    participant User
    participant Client as Claude Desktop
    participant Host as praisonai-mcp
    participant Tools as praisonai.todo.*
    participant Store as todos.json

    User->>Client: "add a todo: Ship it"
    Client->>Host: praisonai.todo.add
    Host->>Tools: TodoTools()._save_todos()
    Tools->>Store: append {id:1, task:"Ship it"}
    User->>Client: "list my todos"
    Client->>Host: praisonai.todo.list
    Host->>Tools: TodoTools()._load_todos()
    Tools->>Store: read todos.json
    Store-->>Tools: [{id:1, task:"Ship it"}]
    Tools-->>Client: "Todo added: 1" then the list
    Client-->>User: "1. Ship it (pending, medium)"
```

***

## Best Practices

> **Workspace switches the visible list**
>
> A todo added inside `~/code/api-service/` writes to `~/code/api-service/todos.json`. Switching directories switches the visible list — useful for keeping unrelated todos apart.

> **Ids are integers under the hood**
>
> The public MCP surface accepts strings, but the underlying id is an integer. `str(id)` is safe; parse ids back as `int`.

> **Delete is safe to retry**
>
> `praisonai.todo.delete` returns `"Todo not found: <id>"` when nothing matched, so a client can repeat a delete without failing hard.

> **Migrate off the old path**
>
> Any tooling that reads `~/.praison/todo.json` is looking at the pre-#4962 layout, which is no longer written. Migrate to `~/.praisonai/todos.json` or the workspace `todos.json`.

***

## Related

> **[Todo CLI](/docs/cli/todo.md)**
>
> Manage todos from the shell.

> **[MCP Memory Tools](/docs/features/mcp-memory-tools.md)**
>
> Four memory tools this host also exposes.

> **[praisonai-mcp Package](/docs/features/praisonai-mcp-package.md)**
>
> The heavy MCP host that registers these tools.

> **[Server: PraisonAI MCP](/docs/deploy/servers/praisonai-mcp.md)**
>
> Deploying the full MCP host.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.