Run analytics tasks on a schedule
The Improve documentation and Analytics report skills also run unattended. A scheduled task in Claude can open a pull request against your documentation every week, or write a report on the past period for your team. The task reaches Kapa through the Kapa connector and loads the skill from Kapa's MCP server, so it needs nothing installed.
Before you start
- A Claude plan with scheduled tasks. See Anthropic's scheduled tasks documentation for the plans and limits.
- For documentation pull requests, your documentation repository on GitHub, connected to Claude.
Connect Kapa to Claude
- In Claude, add a custom connector with the URL
https://mcp.kapa.ai/mcp. - Connect it and sign in with your Kapa account.
The scheduled task acts as you, with your permissions in Kapa.
Create the task
- Create a scheduled task in Claude, for example at
claude.ai/code/routines or with
/schedulein Claude Code. - Choose a weekly schedule. For documentation pull requests, choose your documentation repository.
- Paste one of the prompts below, with your project names.
For documentation pull requests:
Use the Kapa connector. Call get_skill with name "kapa-improve-docs" and follow
it unattended for the Kapa project "<public project name>" over the last 7 days,
using "<internal project name>" as the internal project, which you may search.
Open at most two pull requests and record everything else in
.kapa/docs-improvements.md.
For a report:
Use the Kapa connector. Call get_skill with name "kapa-analytics-report" and
follow it unattended for the Kapa project "<project name>" for the past week,
compared with the week before.
Leave out the internal project if you do not have one.
Verify
Run the task once by hand. A documentation task opens at most two pull requests
on a new branch and commits its decisions to .kapa/docs-improvements.md; a
report task ends with the report. Review any change the pull request marks as
informed by your internal project before you merge it.
Related
- Improve your documentation with an AI agent: the same work, interactively.
- Email reports: the analytics summaries Kapa sends by email.