MCP and AI Skills: connect your Claude assistant to Kalinotes to automate your reviews and documentation

MCP (Model Context Protocol) connectors now let your AI assistant read from and write directly into your Kalinotes projects. Combined with Claude Code skills, they turn note-taking, application review and bug tracking into fully automated workflows — without leaving your terminal or your IDE.

What is the MCP protocol?

The Model Context Protocol (MCP) is an open standard developed by Anthropic that lets AI agents connect to external data sources and tools in a standardised way. In practice, instead of manually copying and pasting content between your assistant and your tools, the MCP connector creates a direct bridge: the AI can read your notes, create new content and update existing entries in real time.

Kalinotes offers its own MCP connector that gives access to your entire workspace: projects, categories, notes and comments. Your AI therefore has the same context as you, with no extra effort.

Setting up the Kalinotes MCP connector

Setup takes three steps from your terminal:

1. Install Claude Code: npm install -g @anthropic-ai/claude-code
2. Retrieve your API token from your Kalinotes account settings
3. Add the MCP server to Claude Code: claude mcp add kalinotes --token YOUR_TOKEN

Once configured, Claude Code has access to all your Kalinotes projects and categories. You can check the connection by asking Claude: "List my Kalinotes projects".

The 8 Claude skills available for Kalinotes

Skills are slash commands (Markdown files placed in ~/.claude/commands/) that extend Claude Code's capabilities with specialised workflows. Kalinotes offers eight dedicated skills, downloadable from the app.kalinotes.com/outils/claude-skills page.

Installing the skills

Each skill comes as a .md file to place in ~/.claude/commands/ for global availability, or in .claude/commands/ at project level for a local scope. Once in place, the corresponding slash command is immediately available in any Claude Code session.

Use case: a complete application review with /kalinotes-qa

Here is a concrete scenario illustrating the power of combining MCP + skills. You have just finished developing a web application and want to run a quality review before delivery, generate bug and improvement notes, view a summary report and share the results with your client or your team.

Step 1 — Connect your AI and install the skills

Once you have set up the MCP connector as described above, download the skills from the Kalinotes tools page. Copy the kalinotes-qa.md and kalinotes-bug.md files into your ~/.claude/commands/ folder. Then open Claude Code in your project directory:

Step 2 — Run the QA review with /kalinotes-qa

Launch the audit by providing the URL of your application (or the local path if it is running in development):

/kalinotes-qa https://mon-app.example.com

Claude Code then goes through the application along four lines of analysis:

• Functionality: checking user journeys, forms, links, navigation
• UX: visual consistency, accessibility, readability of error messages
• Performance: load times, unoptimised images, blocking resources
• Security: missing HTTP headers, exposure of sensitive data, risky configurations

Step 3 — Generate bug and improvement notes

At the end of the audit, Claude automatically creates structured notes in your Kalinotes project. Each anomaly detected generates a bug-type note via /kalinotes-bug, including:

• A description of the observed symptom
• The steps to reproduce the problem
• The estimated severity (blocking, major, minor)
• A suggested fix

The improvement points (UX, performance, accessibility) are recorded in a dedicated "Improvements" category, with a priority note and the expected benefit for the end user.

Step 4 — View the summary report

Once the notes have been created, use the /kalinotes-review skill to generate a consolidated report of the project:

/kalinotes-review

Claude aggregates all of the project's notes and produces a structured summary:

• Number of bugs by severity
• List of prioritised improvements
• Recap of any technical decisions made
• Estimate of the fixing effort

This report can be shared directly with your client or your project manager as a record of the acceptance testing phase.

Step 5 — Share access to the project

Kalinotes lets you share a project with other users from the web interface. Go to the settings of the relevant project and invite your collaborators by email address. You can set two access levels:

• Read only: ideal for a client who follows progress without editing the notes
• Collaborator: full access for a team member taking part in the review

Invited members access the notes created by your AI directly, without needing to install Claude Code or configure the MCP — collaboration stays smooth even for non-developers.

Why this approach is a game changer

Combining MCP + skills eliminates the most time-consuming manual steps of the review cycle: logging bugs, organising notes, writing the report. The AI takes care of the documentation while you focus on decisions and fixes. And since everything is centralised in Kalinotes, the feedback stays accessible, traceable and shareable, whatever the size of the team.