
You can connect ChatGPT to SelfManager.ai through its MCP server, authorize access to your account and use conversation to read or update your tasks.
The basic setup is to add the server in ChatGPT, sign in to SelfManager.ai, install the resulting plugin and select it in a conversation. Then test one read request and one small change.
This guide uses SelfManager.ai throughout, with prompts you can adapt to your own work.
Setup instructions checked October 24, 2026. The steps follow OpenAI’s current custom MCP plugin flow on the web. Account controls and interface labels can change.
Have these ready:
OpenAI’s current instructions use Plugins → Add custom MCP server. Account and workspace policies apply.
The documentation reviewed for this guide does not establish a complete current plan-by-plan eligibility list. Check whether your account exposes the control before buying or changing a ChatGPT plan specifically for this connection.
If you use an organization-managed workspace, an administrator may need to enable the relevant access.
Use the address published on the SelfManager.ai MCP page:
https://api.selfmanager.ai/mcpCopy the complete address, including the final /mcp.
You will enter it in ChatGPT’s server URL field.
In ChatGPT on the web:
This creates the ChatGPT-side connection. Follow any sign-in prompts that appear during setup or installation.
Older instructions may refer to Apps, Connectors or developer mode. The current OpenAI flow checked for this guide uses Plugins.
When the SelfManager.ai authorization screen opens:
For checking your workload, read access is sufficient. Creating or completing tasks requires permission to make changes.
Choose additional access only when it supports the workflow you intend to use.
If you have several SelfManager.ai accounts, confirm the account before authorizing. This avoids a successful connection that appears to contain the wrong tasks.
OpenAI’s current quickstart separates creating a custom connection from installing it.
After creation:
Selecting it explicitly helps make the intended source clear.
You can now ask ChatGPT to use the connected task manager.
Start with a request that does not change anything:
Use SelfManager to show the tables and tasks on today’s date. Do not change anything.
Check the result against the day in SelfManager.ai.
A useful answer should refer to your actual work. If there are no tasks on that date, an empty result may be correct.
You can also ask about a date where you know you have data:
Use SelfManager to show my tables for [specific date]. Read only.
Replace the placeholder before sending.
The purpose is to confirm the connection, account and date before asking for changes.
Once reading works, try a small, recognizable change:
Use SelfManager to create a table for tomorrow called “Connection test” with two tasks: “Check the connection” and “Choose tomorrow’s first priority.” Tell me what was created.
Review any confirmation ChatGPT presents.
Then open tomorrow in SelfManager.ai and check:
A successful explanation in chat is useful, but the actual record is the clearest way to verify the result.
For changes, identify the date, table and task whenever possible.
For example:
In SelfManager, on [date], in the table “Website update,” mark “Check the contact form” as complete.
This is easier to interpret than “Mark that one done,” particularly if several tables contain similar tasks.
If you are unsure which item matches, ask ChatGPT to find and show it first:
Find the task about checking the contact form. Show me the matching table and date before changing anything.
Ask for the change after confirming the right item.
Read today’s SelfManager tables. Suggest a realistic starting order based on the recorded priorities and deadlines. Explain any assumptions. Do not edit the tables.
This separates a recommendation from a change.
You still decide what to do, especially when availability or commitments are missing from the record.
I need to prepare a presentation over the next five working days. The steps are gathering figures, outlining the argument, drafting slides, checking the figures and rehearsing. Suggest how to distribute them. Show me the plan before creating SelfManager tables.
Review the proposed distribution, adjust it, then ask for the approved work to be created.
Specify actual dates when weekends, holidays or a deadline could cause ambiguity.
Add “Send the revised proposal” to my SelfManager table “Client follow-ups” on [date]. Keep the wording exactly as written.
If the table does not exist, tell the assistant whether it should create one or ask you first.
Read my SelfManager work from [start date] through [end date]. Summarize completed work, unfinished items and recorded blockers. Use tracked time where available, and say when information is missing. Do not change anything.
This asks ChatGPT to produce a review from the information it retrieves.
SelfManager.ai’s in-app AI Review is a separate feature. The MCP connection does not launch that internal AI feature on ChatGPT’s behalf.
Check the account and workspace you are using, then review the current ChatGPT controls.
In a managed workspace, ask the administrator whether custom MCP connections are permitted.
A guide for another interface version may show a different route. Use the current controls available to your account.
Confirm that the plugin is installed and selected in the conversation.
Try the direct read-only test again:
Use the connected SelfManager plugin to read today’s tables. Return the actual result.
An answer describing how task management works does not demonstrate that the connection was used.
Check whether you authorized changes in SelfManager.ai.
Additional types of access have their own controls. Review the authorization settings for the action you are attempting.
Use one simple task change to diagnose the problem before trying a larger batch.
Check the date and connected SelfManager.ai account.
Try a specific date where you know a table exists. An empty day is different from a failed connection.
Open the relevant date and table in SelfManager.ai before repeating the request.
If ChatGPT is uncertain, ask it to read the current record. That reduces the chance of creating a second copy of something that already exists.
In SelfManager.ai, open My Profile → Connected apps and disconnect the relevant connection.
You can also remove or uninstall the corresponding plugin in ChatGPT. Revoking access ends the connection; review the task manager directly if you want to change any data already created.
After testing, choose one recurring need.
For example:
Keep requests small enough to check while learning how the connection behaves.
If you are starting with SelfManager.ai, begin your 7-day free trial, add real work for a few days and test the connection with that context.
Yes, through an authorized MCP connection with permission to make changes. Specify the date and table, and verify the result in SelfManager.ai.
This setup uses ChatGPT’s connection controls and SelfManager.ai’s sign-in flow. It does not require you to build a server or write an integration.
Check whether your current account or workspace provides access to adding a custom MCP server. OpenAI’s setup documentation reviewed here does not provide a complete plan-by-plan eligibility matrix, and workspace policies can also affect access.
The connection gives it access to supported tools within the authorization you grant. Ask it to retrieve the relevant dates or tables rather than assuming it has already loaded your entire history.
They are different workflows. ChatGPT can retrieve work and write its own review. SelfManager.ai also has an in-app AI Review feature, which the MCP connection does not run.
The general approach applies when that product provides a compatible MCP server and authentication flow. Use that provider’s official server address and instructions.

Plan smarter, execute faster, achieve more
Create tasks in seconds, generate AI-powered plans, and review progress with intelligent summaries. Perfect for individuals and teams who want to stay organized without complexity.
Get started with your preferred account