Using Binder as an indexing layer for AI agents & docs auto-generation

[IsLand-x]

First of all, great work on Binder! The bidirectional Markdown sync and local-first approach are exactly what I've been looking for in a knowledge base tool.

I have two questions after exploring the project:

1. Can Binder serve as an "indexing layer" for AI agents?

My use case: I want to use Binder as the structured index that my AI agents query when accessing a knowledge base. Is it sufficient to customize the .binder/*.yaml files (types, fields, navigation) to define the schema and indexing structure for this purpose? Or are there additional integration points (e.g., MCP server customization, query APIs) that I should be aware of?

2. Is the docs/ directory self-hosting / auto-generated?

Looking at the docs directory, I see concept documentation (e.g., docs/concepts/changeset.md) that describes Binder's own internals. Are these documents generated automatically by Binder itself ("self-hosting" / "bootstrapping") via the binder docs render command? Or are they manually maintained and simply live alongside the project?

If they are auto-generated, I'd love to understand the workflow: do you author the content as data records in .binder/data/ and then render them out to Markdown? Or is it the other way around — you write Markdown and sync it back into the knowledge graph?

Thanks again for the excellent project. Looking forward to your insights!

[mpazik]

Apologies for the late reply @IsLand-x. I had misconfigured my notification settings.

1. Querying / indexing

Yes. There's CLI, MCP, and HTTP support for search, so customizing .binder/*.yaml (types, fields, navigation) is enough to define your schema and indexing structure. Don't forget binder docs sync or binder docs sync .binder/fields.yaml after. Or ask agent to update schema trough CLI, which is what I usually do.

Agents can query over any of those interfaces. Here's the reference, though it could be clearer: https://github.com/mpazik/Binder/blob/main/docs/concepts/query.md

The full text search is something I am working on though.

2. Is docs/ self-hosting / auto-generated?

Yes. The docs are generated automatically and stay updated on change. Binder is a multi-interface knowledge base, so there are several ways to update it:

• Edit the Markdown file and run binder docs sync [path], or just save the file if you have the VS Code extension installed.
• Use the CLI, e.g. binder update changeset status=draft, or call it from your own scripts.
• Ask an agent to make the changes if you've set up MCP or a skill.

I'd recommend this section: https://github.com/mpazik/Binder#working-with-binder

Binder is my first substantial open-source project and I'm still learning how to explain it and make it easier to set up. If you run into any issues or have feedback, open an issue or start a Discussion. I'd love to hear it. Happy to jump on a call too. Just email me at marek@marekpazik.pl.