Jot adds MCP server to let Claude Code access local notes
A new integration allows AI agents to search, read, and organize your personal notes using the Model Context Protocol.
The short version
- The Jot MCP server allows AI agents to list, search, read, and add new notes to a local folder.
- Notes are saved as Markdown files and stay on your Mac unless the agent sends them to a model.
- The system intentionally excludes a delete tool to prevent AI agents from accidentally erasing data.
- Users can point the tool at custom folders, such as Obsidian vaults, to share notes across apps.
Why it matters
This setup removes the friction of manual context switching by letting your AI assistant 'know' what you've saved throughout the day. It creates a more seamless workflow for developers and researchers who rely on accumulating small bits of information.
Connecting notes to AI
Developers often capture small pieces of information during their workday, such as specific error messages, documentation lines, or ideas from code reviews. Manually pasting these notes into an AI chat to provide context can be a slow and repetitive process. The Jot app for Mac now addresses this by providing a Model Context Protocol (MCP) server.
MCP is a standard that allows AI applications to connect to external data sources and tools. By using this protocol, Jot enables Claude Code, Codex, and Cursor to interact directly with your local notes. This means the AI can pull in the exact context you need based on your previous captures rather than you having to find and paste it yourself.
As sayjot.xyz reported, the integration allows the AI to perform several actions including listing, searching, and reading notes. It can also create new notes directly from the chat interface. This creates a bridge between your personal knowledge base and the AI's reasoning capabilities, making the assistant more aware of your specific projects and history.
How the local server works
The Jot MCP server operates as a local service on macOS, running over stdio. Because it is local, there is no requirement for an account or an API key. The notes do not leave your Mac unless you specifically ask the agent to send them to the model for processing. This architecture prioritizes privacy while still providing the AI with the necessary data to be helpful.
Jot saves each note as a Markdown file in a default folder, typically located at ~/Documents/Jot. When you select text in any application and tap Shift twice, the app saves that snippet. The MCP server watches this folder and updates the app's view in real-time. If the AI adds a new note via a command, it appears in the app's inbox almost immediately because the server performs an atomic write to the file.
The setup involves linking the Jot CLI to your system's local bin and then adding the server to your specific AI client. For example, Claude Code users can add the server using a specific command to make it available across all projects. Other clients like Cursor or Claude Desktop use a JSON configuration file to establish the same connection to the local Jot tools.
Safety and intentional limits
One of the most significant design choices in the Jot MCP server is the deliberate omission of a delete tool. The developer explains that agents can act very quickly and make mistakes. A prompt asking an agent to 'clean up old notes' could result in the permanent loss of files if a delete tool were available to the AI.
To balance utility and safety, the system includes an archive tool instead of a delete tool. Archiving allows the AI to move a note out of the active inbox view while keeping the file safely in the local folder. If an agent archives the wrong note, the user can easily find it and move it back, ensuring that no data is lost due to a misunderstanding of a prompt.
Users still retain full control over deletion through the main Jot app. Deleting a note in the app is a single action that first moves the file to a trash folder. This distinction ensures that while the AI can help organize and manage the lifecycle of your notes, the ultimate authority over data destruction remains with the human user.
Advanced usage and workflows
The integration is designed to work with plain language prompts rather than requiring specific technical commands. Users can ask the AI to summarize notes about a specific migration, list todos tagged for a certain repository, or read a specific documentation snippet to apply a fix to a file. The AI automatically selects the correct tool based on the intent of the request.
For more precise results, the developer suggests using short, specific words for searches because the system matches substrings. Additionally, since the standard search only looks at open notes, users can specifically request to see archived notes by asking the AI to set the archived parameter to true. This allows for a nuanced way of querying both current and historical information.
The tool is also flexible enough to point at different directories. By using environment variables or specific settings, users can point the MCP server at an Obsidian vault. This allows the AI to read and interact with the same notes that the user sees in their Obsidian workspace, effectively extending the AI's reach into existing knowledge management systems.
Availability and pricing
Jot is compatible with macOS 12 or later on both Apple silicon and Intel chips. The app provides a three-day free trial for new users to test the functionality before committing to a purchase. The pricing model is a one-time fee of $29 for a license for one Mac.
The developer is offering a limited-time discount for early adopters, with the first 50 one-Mac licenses available for $9. This price includes the MCP server, which is necessary for connecting the app to Claude Code, Codex, and other supported AI clients. The full tool reference for the MCP server is available on GitHub for those who want to see the underlying technical specifications.
For users who do not want to use the MCP protocol, the folder structure still remains functional. Jot writes two specific files—INDEX.md and SKILL.md—into the folder to help agents that only have file-reading capabilities. These files explain the layout and provide instructions on how the agent should interact with the notes, ensuring some level of compatibility across different AI tools.
Sources and links
- Codegithub.com/MohakBajaj/jot-mcp github.com
- Reportingsayjot.xyz: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by sayjot.xyz · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
