MCP lets you connect SingularityApp to an AI service and work with your tasks directly from the chat.
Once connected, the AI assistant can read data from SingularityApp, help you review projects and plans, and, if you grant the required permissions, create and edit tasks, projects, tags, habits, Kanban boards, and other items.
MCP is not a separate AI assistant inside SingularityApp. You continue working in your usual AI client—ChatGPT, Codex, Cursor, or Claude—while MCP connects the AI service to your SingularityApp account.
What you can do after connecting MCP
MCP is useful for more than commands like “create a task.”
The AI gets context from your planning system and can work with several types of SingularityApp data at once.
For example, you can ask:
“How many projects do I have right now?”
“Show me my overdue tasks and group them by project.”
“Review the ‘Get a Driver’s License’ project and tell me what the most important next action is.”
“Create a project called ‘Prepare for the Move’ and add these five tasks to it.”
“Check whether this task already exists before creating a new one.”
“Collect all my book ideas into one note and archive the old separate tasks.”
“Review my projects and help me understand which ones are actually active and which have turned into idea storage.”
“Review my tasks without dates and suggest which ones should actually be scheduled.”
“Move these tasks to another project.”
“Show me my active habits and find duplicates.”
The main difference from a regular AI chat is that you don’t need to copy your task list into the conversation manually or keep it up to date yourself. The AI can retrieve the data directly from SingularityApp.
Before connecting
Use the following SingularityApp MCP server address:
https://mcp.singularity-app.com/mcp
When you connect for the first time, the SingularityApp authorization page will open. Sign in to your account and confirm access.
After that, the AI app will appear in the list of connected apps in your SingularityApp account.
If other MCP servers are already configured in the file, add singularity to the existing object without removing the other connections.
Field names and the location of the configuration file may vary. For example, VS Code uses the servers object instead of mcpServers. Before editing the file, check the instructions for your AI client.
What data AI can access
By default, MCP gives the AI access to the main SingularityApp tools.
Depending on the selected toolset, the assistant can work with:
Tool
Access rights
Tasks (tasks)
read tasks and checklists;
create and edit tasks;
change task completion status.
Projects (projects)
read projects;
create and edit projects.
Core entities (meta)
retrieve related SingularityApp data;
create and edit available entities.
Habits (habits)
read habits;
create and edit habits.
Kanban boards (kanban)
read boards;
create and edit boards.
Tags (tags)
read tags;
create and edit tags.
Batch command execution — Batch (system) — is disabled by default.
Without Batch, the AI makes changes one at a time. For example, if you ask it to edit ten tasks, it will update them sequentially.
This is safer: if the AI misunderstands your request, it’s easier to stop the process before a large number of items are affected.
You can enable Batch separately if you want to allow the AI to perform up to 20 operations in a single request.
Learn more about toolsets and the toolsets parameter in the How to limit AI access section below.
Connecting to ChatGPT
In ChatGPT, MCP can be used through apps and plugins that support remote MCP tools.
Full MCP access, especially for creating and editing data, depends on your ChatGPT plan and workspace settings.
First, enable Developer mode if it isn’t already enabled. After that, you’ll be able to add a custom MCP connection in the apps section, if this feature is available for your account.
Use this address to connect:
https://mcp.singularity-app.com/mcp
Step 1. Open Apps or Plugins
Open ChatGPT settings and find the Apps / Plugins section.
If your interface allows you to add a custom connection, choose the option to create a new app or MCP connection.
Step 2. Add SingularityApp
Enter this address:
https://mcp.singularity-app.com/mcp
If ChatGPT asks you to enter a connection name, you can use:
SingularityApp
Save the connection.
Step 3. Sign in to SingularityApp
ChatGPT will redirect you to the authorization page.
Sign in to your SingularityApp account and confirm access.
After authorization is complete, return to ChatGPT.
Step 4. Check the connection
Open a new chat and ask a simple question that requires data from SingularityApp.
For example:
How many projects do I have in SingularityApp?
or:
Show me my active projects.
If ChatGPT asks you to select or allow SingularityApp for this request, confirm it.
If there is no option to add MCP
Check whether Developer mode is enabled in ChatGPT.
To do this:
Open Settings.
Go to Apps.
Find Developer mode.
Turn it on.
Return to the apps section and check whether the option to add a custom MCP connection is now available.
If you don’t see the Developer mode setting, custom MCP connections may not be available on your ChatGPT plan or may have been disabled by your workspace administrator.
Connecting to Codex
Codex can connect directly to remote MCP servers.
You can configure the connection through the command line, the Codex app, or an IDE extension. MCP settings are saved in the Codex configuration and can be used by its local clients.
If you have already connected SingularityApp in ChatGPT
If SingularityApp is already installed in ChatGPT as a plugin and you use Codex with the same ChatGPT account, you usually don’t need to add the connection again. Installed plugins and their MCP tools are available in both ChatGPT and Codex.
Open a new task in Codex and check whether SingularityApp appears among the available plugins.
If SingularityApp doesn’t appear, or if you use Codex CLI or an IDE extension with a local configuration, connect the server using one of the methods below.
Your browser will open. Sign in to SingularityApp and confirm access.
Check that the server was added
Run:
codex mcp list
The singularity server should appear in the list.
In the Codex interface, you can also use:
/mcp
This command shows the connected MCP servers.
Option 2. Connect through the Codex interface
If you use the Codex app:
Open Settings.
Go to MCP servers.
Click Add server.
Select Streamable HTTP.
Enter SingularityApp as the server name.
Enter this URL:
https://mcp.singularity-app.com/mcp
Save the server.
If you see an Authenticate button, click it and sign in to SingularityApp.
Restart the app if Codex asks you to do so.
Option 3. Connect in the Codex IDE extension
In the Codex extension:
Open the settings menu.
Select MCP servers.
Click Add server.
Select Streamable HTTP.
Enter the SingularityApp MCP address.
Save the settings.
Restart the extension.
If the server requires authorization, click Authenticate.
Codex CLI and the IDE extension use the same MCP configuration on the same host, so you usually don’t need to add the server again after configuring it once.
Check the connection
After authorization, ask Codex:
Show me my projects in SingularityApp.
Then test an action that changes data:
Create a test task called “Check MCP” without a date.
After testing, you can delete the task or move it to the archive.
Connecting to Cursor
Cursor supports remote MCP servers over HTTP and OAuth. The Cursor app and Cursor CLI use the same MCP configuration, so you only need to add the server once.
Option 1. Connect through the Cursor app
For a personal connection that will be available across all projects, open or create this file:
If the file already contains other MCP servers, add singularity to the existing mcpServers object without removing the other settings.
Save the file and restart Cursor.
After restarting Cursor, open Customize in the sidebar and select MCPs.
Find the singularity server and make sure it’s enabled. If Cursor asks you to authorize the connection, open the OAuth window, sign in to SingularityApp, and confirm access.
Option 2. Connect via Cursor CLI
Cursor CLI uses the same mcp.json file as the Cursor app. First, add the configuration from the previous option, then open a terminal.
Check that Cursor CLI can see the server:
agent mcp list
If the server is disabled, enable it:
agent mcp enable singularity
Start OAuth authorization:
agent mcp login singularity
Cursor CLI will open your browser. Sign in to SingularityApp and confirm access. After authorization is complete, the server will be available in Cursor CLI.
The Cursor app will continue using the same configuration. If it asks you to sign in separately, complete OAuth authorization in the app.
To check the available tools, run:
agent mcp list-tools singularity
Check the connection
Open a Cursor Agent chat and ask:
Show me my active projects in SingularityApp.
Cursor will automatically use available MCP tools when they are relevant to your request. By default, it may ask for confirmation before calling a tool.
You can connect SingularityApp through Claude Code in the terminal or by adding a remote custom connector in Claude Desktop.
Option 1. Connect via Claude Code
Open a terminal and run:
claude mcp add --transport http singularity https://mcp.singularity-app.com/mcp
This command adds the remote HTTP server to your Claude Code configuration.
Check the connection status:
claude mcp list
If the server requires authorization, start the OAuth flow from the terminal:
claude mcp login singularity
Your browser will open. Sign in to SingularityApp and confirm access.
You can also authorize the connection from an interactive Claude Code session:
Start Claude Code with the claude command.
Enter /mcp.
Select the singularity server.
Choose the authorization option and complete the sign-in process in your browser.
Option 2. Connect through Claude Desktop
SingularityApp can be connected to Claude Desktop as a remote custom connector.
For an individual account:
Open Customize.
Go to Connectors.
Click + and select Add custom connector.
Enter this URL:
https://mcp.singularity-app.com/mcp
Click Add.
If Claude opens the OAuth authorization page, sign in to SingularityApp and confirm access.
Configure tool permissions
After connecting SingularityApp, choose which actions Claude can perform without asking for additional confirmation.
Open Settings → Connectors.
Select SingularityApp.
Under Tool permissions, configure permissions for tool groups or individual actions.
Four options are available:
Always allow — Claude can use the tool without asking for confirmation.
Needs approval — Claude will ask for your permission in the chat before using the tool.
Blocked — Claude won’t be able to use the tool.
Custom — permissions can be configured separately for each tool.
If you leave Needs approval enabled, Claude will ask for confirmation every time it uses the corresponding tools. This is normal behavior and doesn’t mean there is a connection problem.
For a simpler, safer setup, you can allow read-only tools without confirmation and keep Needs approval for actions that create, edit, or delete data. You can block any tools you don’t need.
You can apply a permission to an entire group, for example all read-only tools, or configure each action individually.
In Team and Enterprise accounts, some permissions may be restricted by your organization administrator. In this case, you won’t be able to select a less restrictive permission level yourself.
In Team and Enterprise workspaces, users may not be able to add a connector themselves. An Owner or Primary Owner must first add it under Organization settings → Connectors → Add → Custom → Web. After that, the user can open Customize → Connectors and click Connect.
To use SingularityApp in a specific conversation, click + to the left of the message field, open Connectors, and enable the added connector.
Check the connection
Ask Claude a simple read-only question:
What projects do I currently have in SingularityApp?
If Claude asks for permission to use a tool, review the action and confirm it.
Gemini Spark is an AI agent in the Gemini app. You can connect custom apps to it through remote MCP servers, then ask Spark to work with data from SingularityApp.
You can set up the connection in the web version of Gemini at gemini.google.com.
Before you connect
To use Gemini Spark, you need:
a personal Google Account — work and school accounts are not supported yet;
to be 18 or older;
a Google AI Pro or Ultra subscription;
Keep Activity turned on.
Spark and custom connected apps may also be available only in certain countries, interface languages, and stages of the feature rollout. If you don’t see the Spark tab or the Custom apps section, check your plan and whether the feature is available for your account.
You can also connect SingularityApp to other AI clients that support remote MCP servers and OAuth.
Interfaces and access requirements in these services may change, so the sections below provide only brief guidance and links to the official documentation.
Perplexity
Perplexity supports custom remote MCP connectors with OAuth authorization.
When setting up the connection, choose a remote connector, enter SingularityApp as the name, use https://mcp.singularity-app.com/mcp as the server address, select Streamable HTTP as the transport, and OAuth as the authorization method.
Availability depends on your plan and organization settings. We haven’t tested the SingularityApp connection with Perplexity yet, because verification requires access to the relevant Enterprise features.
VS Code with GitHub Copilot supports remote MCP servers. Open the Command Palette, run MCP: Add Server, select a remote HTTP server, and enter the SingularityApp address. After saving the connection, start authorization from the mcp.json file if VS Code asks you to do so.
GitHub Copilot CLI also lets you add the server from the terminal:
If direct MCP connection isn’t available in the chat
Not every AI service lets you add a third-party MCP server by URL directly in its standard chat interface. In some cases, you can connect the model to SingularityApp only through an API, a cloud platform, or a custom agent.
If you’re building your own integration directly through an API, use the SingularityApp REST API.
DeepSeek
The DeepSeek chat currently doesn’t support direct connections to custom MCP servers.
You can use DeepSeek models through the API or select them as the model in a compatible agent tool. To work with SingularityApp, you’ll need to create a DeepSeek API key and configure an external MCP client or integration layer between the model and the MCP server yourself.
This is a technical setup intended for developers. You can’t simply paste the SingularityApp MCP server address into the standard DeepSeek chat.
After connecting, you don’t have to give the AI only short technical commands. You can describe the task just as you would to an assistant.
For example:
Review my projects in SingularityApp.
Don’t change anything yet.
Find projects without a clear outcome, duplicate projects, and projects that look more like reference lists or wish lists.
Suggest which projects should stay active, which should be turned into notes, and which can be archived.
The AI will first retrieve the existing data, analyze it, and only then suggest changes.
Another example:
Review my overdue tasks.
Don’t reschedule them automatically.
First, show me the list and ask which tasks are already completed, which are still relevant, and which can be deleted or archived.
Or:
I want to create a project called “Bicycle.”
First, check whether a project with the same name or a similar topic already exists.
If it does, show it to me and suggest using the existing project. If not, create a new one.
This approach is especially useful if you have a large number of tasks. The AI can work not only as a quick input tool, but also as an assistant for reviewing and organizing your planning system.
A useful rule: check first, then make changes
AI can work with large amounts of data faster than a person, but it understands what you want from the instructions you give it.
For bulk changes, it’s useful to be explicit:
Show me what you’re going to change first. Don’t make any changes without my confirmation.
Or:
Before creating a task, check whether a duplicate already exists.
Or:
Don’t assign dates automatically. Suggest an option first.
These instructions are especially useful when reviewing old projects and overdue tasks.
How to limit AI access
You don’t have to give the AI access to all SingularityApp tools.
You can define which tool groups are available using the toolsets parameter in the URL.
The system toolset enables Batch command execution.
With Batch enabled, the AI can perform up to 20 operations in a single request. For example, it can update several tasks at once.
For your first connection, we recommend keeping Batch disabled.
Start with regular sequential operations and make sure the AI interprets your requests correctly. Enable Batch later only if you actually need bulk operations.
How to revoke access
You can revoke an AI app’s access to SingularityApp at any time.
Find the connection you want to remove and revoke access.
After that, the app will no longer be able to access your SingularityApp account using that authorization.
What to try after connecting
Start with safe read-only requests:
What projects do I currently have?
Which tasks are overdue?
Show me tasks without a project.
Then try a small change:
Create a task called “Test MCP” without a date.
After that, you can move on to more advanced scenarios, such as reviewing projects, planning your week, finding duplicates, working with Kanban boards, and cleaning up accumulated tasks.
MCP doesn’t replace your planning system. It gives AI a way to work directly with your planning system.
The documentation is automatically generated. If you haven’t found the answer to your question, please contact us at support..
