I gave Claude Code entry to my note-taking instrument and stopped doing every part in my workspace manually
Claude Code is superb at working with information, however my tasks hardly ever start inside a core repository. Requirements, analysis, troubleshooting notes, article concepts, and half-formed plans often reside in AFFiNE, my observe taker of alternative.
This separation creates an annoying hole. Claude might perceive the code in entrance of it, however it knew nothing in regards to the choices, experiments, and background info saved elsewhere. So I finished repeatedly copying notes into the terminal and gave Claude Code direct access to my AFFiNE workspace.
Copy, paste, repeat, and now three variations of the reality
CLAUDE.md handles the venture, not the entire data base
Before this setup, utilizing Claude Code alongside AFFiNE concerned much more guide work than it ought to have. If I wished Claude to comply with a top level view, I copied the related part from AFFiNE and pasted it into the immediate. If Claude produced a helpful analysis abstract or implementation plan, I copied that again right into a observe.
This labored, however it additionally created duplicate info. The observe in AFFiNE would say one factor, a Markdown file within the venture would comprise a barely newer model, and the Claude Code dialog can be in a totally completely different area.
Claude Code already offers persistent project instructions by way of information like CLAUDE.md, together with an computerized reminiscence system. However, these mechanisms are primarily tied to a venture or native machine quite than serving as a substitute for a broader data base. I nonetheless wished AFFiNE to stay the house for all my notes.
MCP beats logging right into a browser Claude does not totally management
Read-only by default, scoped, expiring, revocable
There are two most important methods to attach Claude Code to AFFiNE. The first is through an MCP server the place you need to use both AFFiNE’s built-in MCP or third-party community made ones from GitHub. The different method is to log into your AFFiNE workspace from Claude Code’s built-in browser, however that may expose your AFFiNE credentials in a browser you do not totally management and is not as dependable as an MCP connection.
Security is a significant concern right here. An MCP connection can expose delicate notes and doubtlessly modify them, so it ought to use restricted, recoverable credentials. AFFiNE’s MCP implementation helps workspace-scoped credentials, expiration, and read-only entry by default, making it doable to start out cautiously earlier than enabling broader permissions. Anthropic too recommends verifying that you trust an MCP server before connecting it.
Retrieval, not automation, was the precise unlock
A obscure query now pulls precisely the precise web page
The instant profit publish connection wasn’t automated note-taking. It was retrieval.
Instead of discovering a doc myself and pasting a number of paragraphs into the terminal, I might inform Claude to look AFFiNE for my notes a few specific venture. Claude might find the related pages, learn them, and use that materials whereas engaged on the code.
That made obscure requests surprisingly efficient. I might ask Claude one thing random like checking my ESP32 sensor notes and proceed with the firmware plan and it might work. Since I additionally keep a list of all my sensors and microcontrollers, Claude can now question it and counsel related components for a brand new venture from what I’ve readily available.
This is way extra helpful than stuffing every part into one huge CLAUDE.md file. Anthropic notes that these information load into the context window when a session begins, so keeping them focused improves adherence and avoids spending context on info that won’t matter to the present process. With AFFiNE related, Claude might fetch detailed info solely when it wanted it.
Reading works as we speak, writing remains to be catching up
The built-in MCP server AFFiNE has works nice for learn entry, however write entry remains to be rolling out progressively. At the second, it is nice for looking out or studying your notes, quite than have Claude create or edit paperwork.
One resolution is to make use of a neighborhood server obtainable on GitHub that talks to AFFiNE’s collaboration APIs over stdio or HTTP. But utilizing it means trusting third-party code together with your workspace credentials and working/sustaining it your self.
Now I do not need Claude meddling in my notes, but when I make a discovery or repair a very annoying bug, I’d just like the findings to be recorded in the identical place as the remainder of my notes. Currently, I’m fixing this by having Claude document every part in markdown information after which importing it into AFFiNE manually.
You can keep away from this by utilizing AFFiNE in Claude Code’s inbuilt browser. But meaning exposing your workspace credentials in a browser you do not totally management.
AFFiNE stayed the pocket book, Claude simply realized to learn it
Notes keep editable by hand, Claude simply fetches on demand
This setup will not be an alternative choice to Claude Code’s native venture reminiscence. I nonetheless use CLAUDE.md for repository-specific guidelines, instructions, and coding conventions. AFFiNE handles the bigger physique of data: analysis, venture historical past, concepts, check outcomes, and knowledge shared throughout repositories.
The largest enchancment is that I now not keep an AI-friendly shadow copy of my workspace. My notes stay in AFFiNE, the place I can set up and edit them usually, whereas Claude Code can retrieve or replace them when the duty requires it. The AI didn’t exchange my note-taking system, it simply grew to become part of it.

