Nifty's MCP server connects your assistant to your workspace. It reads your projects, tasks, and docs — and, with your permission, updates them on your behalf, all while respecting every permission you already have.
In short: you tell your assistant what you want, and it does the clicking in Nifty for you.
Here's the problem it solves. The moment AI starts doing the work, things drift, your assistant creates something in a chat window, your teammate's assistant does something else, and nobody's board reflects reality. Nifty becomes the one source of truth your whole team, human and AI, works from.
What you can do
Just tell your assistant what you want:
Get instant status. "Summarize where the Q3 launch stands and list what's blocked."
Fill in your board. "Create tasks in the Marketing project for each action item above, assign owners, and set due dates."
Draft and file docs. "Write a project brief for the website redesign and save it as a Doc in that project."
Catch what's slipping. "Find tasks across my projects with no assignee or due date."
Plus, your assistant can post updates into a project discussion or DM a teammate directly, so communication lives alongside the work.
Even better: your AI connects everything
If your assistant already talks to your other tools over MCP, it can read from them and build the results straight into Nifty, as real tasks, projects, docs, priorities, and assignees:
Prevent churn. "Check PostHog for accounts trending toward churn and create outreach tasks in Nifty, assigned to their account owner."
Triage support. "Turn today's support tickets into tasks on the Bugs board, set priority by severity."
Track engineering. "Take the open GitHub issues and file them in the Sprint project with owners and due dates."
Your AI does the reading. Nifty holds the work — structured, assigned, and visible to your whole team.
Why it's built for AI
Fast and token-efficient. Nifty's MCP server is AI-native, so each request uses as few tokens as possible.
Broad and growing. Work across Portfolios, Projects, Lists, Tasks, Docs, Goals, Messages, Forms, and more, with new capabilities added regularly.
Easy to install. For most assistants it's click-to-connect. No config files or scripts to manage.
You stay in control
Your assistant only ever does what you permit:
Control | What it means |
Off by default | MCP is opt-in. Nothing is shared until you create a token or authorize a connection. |
It connects as you | Your assistant uses a scoped access token you generate — never your password — and can never go beyond your own role permissions. |
You choose the scope | Pick Read-only (default and recommended), Full access, or Custom for the exact actions the assistant may take. |
Set an expiry, revoke anytime | Give tokens a lifetime of up to 365 days and revoke any of them whenever you want. Access stops within about a minute. |
Your data, your choice of provider | In-scope data is shared with the AI provider you connect, under their terms and privacy policy — so connect assistants you trust. |
What your assistant can access
Nifty's MCP scope covers your whole workspace:
Work structure
Portfolios → Projects → Lists, Statuses, Tasks
The full project hierarchy, your assistant can navigate and organize across all of it.
Task details
Tags · Dates · Dependencies · Custom Fields
Everything that lives on a task, so your assistant can set priorities, deadlines, and relationships.
Content & collaboration
Docs · Messages · Goals · Forms
Create, update, and read docs, send messages and DMs, and set up forms for intake, so your assistant works where your team already talks.
People
Members
So your assistant can assign owners, tag teammates, and route work to the right person.
Setting up your assistant
Setup varies slightly by assistant, but every install starts in the same place: open the Settings option in your profile dropdown, then go to the MCP & AI Assistants tab in the sidebar.
App Scopes
When integrating with the MCP, you'll be prompted to select a scope based on your needs. The below will help you understand the different scope types.
Read only
Can only retrieve/view data (list, get, search)
Cannot create, update, or delete anything
Good for: dashboards, reporting, auditing, or giving your LLM visibility without risk of changes
Full Access
Can read and write — create, update, delete, and retrieve anything the MCP server supports
No restrictions on what actions can be taken
Good for: power users who want their LLM to take full action on their behalf
Custom
You select which specific permissions to grant
Best when you know exactly what the token will be used for
Good for: situations where you want your LLM to do some things but not others — e.g. create tasks but never delete projects
Claude Desktop via OAuth
Follow the below steps to add a custom connector for the Nifty MCP in Claude.
Start by opening your Claude desktop app and go to Settings within your Profile.
Next, navigate to Connectors in the settings sidebar and Add a new custom connector.
Copy the MCP Server and Client ID details from step one in Nifty as seen below.
Add these details into the Add custom connector pop up then Add the connector.
Following this step, you will be requested to log into Nifty and approve Claude's access. Restart the Claude desktop app and give the Claude a prompt such as "What's the latest task updated in Nifty".
Cursor / Custom
Cursor / Custom implementations will require you to add a small amount of code into the config file of your LLM, but first, you'll need to create a token.
Add Your Code to the Config
Copy the code snippet with the copy icon to the right of the text box.
Cursor Config
From inside of the Cursor desktop app, select Settings. In the Settings sidebar, select Tools & MCPs and Add Custom MCP.
A sidebar editor will open for you to paste the code snippet into the config file. Note that the {mcpServers:} code may already be there, meaning you can remove the duplicate of this code when pasting the code snippet from Nifty.
Save the added code and restart the Cursor desktop app.
Final Steps
Back in Nifty, select I pasted it.
Restart your LLM and send a prompt for the Nifty MCP, after which you can Test your connection to confirm that your token was properly added into the config.
The two most common errors faced in this step are either
Code added to the incorrect place of the config file
Failure to restart your AI client application following the code addition to the config file
In addition to this test, you can try prompting your LLM to interact with Nifty via the MCP, at which point your AI assistant will verify if it is able to reach the Nifty MCP server.
OpenAI Custom App via OAuth
For the MCP beta, you'll need to create an OAuth app in ChatGPT.
Creating a ChatGPT app in Nifty
Copy the URL in Step 1 of the ChatGPT connection steps, then go to the App Center in Nifty.
Navigate to the Integrate with API tab and Create app.
Name the app ChatGPT and add the Redirect URI copied from before. Select your scope and Create app.
Be sure to save your Client ID and Client Secret, as you will need them for the next step in creating the ChatGPT app.
Creating a Nifty App in ChatGPT Browser
Start by navigating to the Profile of ChatGPT in the browser app and go to Settings.
In the Apps section of Settings, go to Advanced settings.
Enable Developer mode, then select Create app.
In the New App screen, title the app Nifty and place the copied URI in the Connection field. Click the tickbox, and click on Advanced OAuth settings.
In advanced OAuth settings, input your Client ID and Client Secret below the Callback URL, then Create the app which will prompt you to sign into Nifty.
After the app is created, log into Nifty via the ChatGPT interface to complete the integration.