Your AI client, signed in as you.
Connect Claude, ChatGPT, Cursor or Gemini to your plnnrly organisation, or call the REST API from your own code. Both act as you, with your permissions.
Building an integration? Start with the REST API
Type the address you sign in to plnnrly on, such as acme.plnnrly.com. The AI panel in plnnrly (the sparkles icon in the top bar) shows the exact URL.
Where do I find my URL?
Type the address you sign in to plnnrly on, such as acme.plnnrly.com. The AI panel in plnnrly (the sparkles icon in the top bar) shows the exact URL.
Then add it to your client
A custom connector on your Claude account. It then works in Claude Desktop and the mobile apps too.
- In Claude, open Customize › Connectors and choose Add custom connector.
- Enter
plnnrlyas the name and paste the server URL. - Click Add, then sign in to plnnrly and click Approve.
The same connector as on claude.ai. If you added it there, it is already here.
- In Claude Desktop, open Customize › Connectors and choose Add custom connector.
- Enter
plnnrlyas the name and paste the server URL. - Click Add. A plnnrly window opens in your browser: sign in and click Approve.
Run this in your terminal, then sign in with claude mcp login plnnrly.
claude mcp add --transport http plnnrly https://<your-url>/mcp/v1ChatGPT on the web adds plnnrly as a custom MCP server.
- Open chatgpt.com/plugins and choose +, then Add custom MCP server.
- Paste the server URL under Connection and choose OAuth.
- Create and install the plugin, then sign in to plnnrly and click Approve.
Add this entry to ~/.cursor/mcp.json. Cursor asks you to sign in to plnnrly when it connects.
{
"mcpServers": {
"plnnrly": {
"url": "https://<your-url>/mcp/v1"
}
}
}Run this in your terminal, then start gemini and sign in with /mcp auth plnnrly.
gemini mcp add --transport http --scope user plnnrly https://<your-url>/mcp/v1What happens when you ask
- You ask
Which of my work items are still open?
- Tool
api_read- Request
- GET
/work-itemsfilter[assignee_id]=mefilter[resolved]=false - Checked
- Token scope api:
read: project_management - Permission project_management::
view_backlog filter[resolved]also needs workflows::read and api:read: workflows
- Token scope api:
- Answer
- 200 OKYour open work items, a page at a time
The same checks as the REST API. A client never does more than you can.
What the server can do
There is no tool per feature: a client searches for the operation it needs, reads what it accepts, then calls it as you. Everything except file uploads is in reach.
- search_api
Finds the API operation for a task and shows what it accepts.
Reads the docs, not your data
- api_read
Reads anything the REST API can read.
Your read permissions
- api_write
Creates and changes data. Deletes only when your organisation allows it and you confirm.
Your write permissions; asks before deleting
- get-current-user-tool
Returns your own profile.
Your own profile
Before you connect
MCP is off for new organisations. An administrator switches it on, and it only works while the REST API is on.
| Setting | Default | What it does |
|---|---|---|
| Enable REST API | On | Lets people call the REST API with their own API tokens, and lets apps sign them in with OAuth. |
| Enable MCP | Off | Lets people connect an AI client, such as Claude or ChatGPT, by signing in. The client works through the REST API, as the person who connected it.Needs Enable REST API. |
| Allow destructive actions | Off | Lets a connected AI client delete what the REST API can delete, within the person's own permissions and only after they confirm.Needs Enable MCP. |
Settings › Administration › MCP & API
For administratorsPrefer code?
The REST API reads and changes projects, work items, hours, Drive documents, customers, invoices, leave and planning. Every operation states the token scope and the permission it needs.
- Create a token: user menu › API access › Create token.
- Find your base URL:
https://<your-url>/api/v2/. - Make your first call:
GET /api/v2/shows what your organisation exposes.
The reference covers 228 operations, generated from plnnrly's code on every release.
curl https://<your-url>/api/v2/ \
-H "Authorization: Bearer <token>"Guides
Start
- MCP overview
How connecting works, and what a client can do once connected.
- For administrators
The three switches, and who may create tokens and connect apps.
- Getting started
Create a token, find your address, then make your first reads and your first write.
Working with data
- What you can ask
Example questions per area, with the operation each one uses.
- Pagination
Cursor paging by default, page numbers where a collection counts its rows.
- Filtering and sorting
Bracket filters with operators, free-text search and sort order.
When things go wrong
- Troubleshooting
When a client cannot connect, cannot sign in or has a request refused.
- Errors and retries
Every failure is a problem document. Act on its type, not on the status code alone.
- Rate limits
What each request costs, and the headers that show how much is left.