Capacities MCP server

Trying to decide whether to move from Tana to Capabilities and this is the main thing for me as I operate mainly from Claude Cowork

Hello everyone! with [release 62](https://capacities.io/whats-new/release-62) we launched the first experimental version of AI Connectors.

If there's any apps you'd like the MCP to connect to that are not currently available, please create a ticket for that specific app to connect to our MCP server.

We'll continue developing the MCP server and adding more capabilities to it in line with our [API work](https://feedback.capacities.io/b/1v35e680/feature-ideas/expanded-api-endpoints).

Link to the docs: https://docs.capacities.io/reference/ai-chat-connectors

The OpenAI connector seems to be for connecting to the web version but not Codex. Maybe it will work with the custom MCP steps but instead I just removed the web connection.

The experience connecting Capacities via MCP with other tools than ChatGPT, Claude, or Curser is not clear to me as a non-technical user.

I'll just give an example, how easy it is with Todoist:

I use TypingMind as a ChatUI and to connect via MCP all I usually have to do is (see screenshot attached):

1. enter the MCP Server URL
2. Enable OAuth authentication
3. Connect
4. Enter OAuth credentials
5. Done

This doesn't work with capacities. I see in the Capacities documentation, that there are more steps to connect to custom tools, but a) why do other tools get this all done automatically and b) the tool that I use doesn't even have the option to enter all those things that I read in the capacities MCP docs.

I hope this get's easier, but I'm glad that MCP is happening :)

So I had Claude Code write a Capacities MCP server and it works. Unfortunately, Capacities API doesn't have any way to actually read notes. Only fetching titles. Sooooo close

Can you please consider including a "fetch_weblink_content" tool so that we can use the MCP to search for specific information from weblinks stored in Capacities? Thanks!

Yes!

How much I dream of seeing a full-fledged interaction of AI Agents with Capacities. This is the acceleration of work with the second brain that I've been waiting for all my life.

Yes, Must have given the complexity of the Capacities interface particularly for new users.

Must have feature

Yes! I have been using Readwise MCP within Claude and it's amazing being able to retrieve my highlights for research. It would be great to have the same capability with Capacities - and, even better, the ability to create new items in Capacities from Claude.

Oh heck yes! Would be huge

This is essential.

This is essential missing functionality, at least for me, in 2026.

This would be a huge one, it really would cement capacities as an invaluable tool in my workflow.

I also agree with the idea. I would be fine if all we have is a more powerful set of APIs so that we can build our own, or adjust an existing MCP server to connect to Capacities.

Seria incrível rodar todas as minhas notas dentro do Claude!

I'm super excited about this. I came to the Feedback board to suggest it. I'm using the Capacities <> Notion MCP (we use Notion in my workplace) and it's incredibly helpful. I'd love the same functionality with Capacities!

I love that this is planned!

I just hope there will be some kind of fair use policy so that people don't just start pushing loads of data into Capacities like it's nothing and use so much storage, that the Cpacities team would be forced to make capacities more expensive.

I'm surprised this hasn't gotten more attention and commentary. I 100% agree, and am glad to see the "planned" tag on the original post. With Claude Code, Claude Cowork, and whatever Gemini, ChatGPT and Grok come up with, this seems enormously valuable to me. There's a good post on reddit that summarizes how useful this will be -- and how important the author thinks it is for Capacities. I tend to agree, but was surprised by the aggressive pushback by some redditors. Hope this gets traction and is prioritized.

Ability to create objects, object types, and bulk edit them! SO for example i have a research with 12 topics, so i let ai through mcp enter each of 12 objects and create a small report on the topic, so when i start all topics have some base to work on. Or i ask ai to find me the untagged objects and tag them for me based on my library. or help me to tidy up some stuff!