Picture this. You're working on a

project in Claude. Then context fills

up. So you start a new conversation.

Then all of a sudden the agent doesn't

know what to do next. Sure, you can add

a markdown file, but for large projects

that can take up a lot of context, and

it's easy for the agent to mess up

priorities. Beads aims to fix all of

that by giving your agent an issue

tracker that can be version controlled

and can even be synced to Jira. But why

would anyone use this over SpecKit or

Kira? Let's find out. And before we do,

don't forget to hit subscribe.

Beads was created by the ex head of

engineering at SourceCraft, the same

company that created AMP, as a way to

chain issues together so that agents can

perform complex tasks in the right

order, which is what we're going to do

with this project and app to help me

view my brew packages. After setting up

beads in the project, we should have a

beads folder which contains an SQLite

database for all of our issues along

with some other database files, our

issues in a JSON lines format, the beads

demon to autoexport database changes to

JSONL for git compatibility, and the

socket enables storage and retrieval of

issues via CLI commands. We'll talk more

about CLI commands a bit later, but

basically for the other files, they work

in a really cool way. Take a look at

this. So because only JSONL files are

committed to the git repo, not the

SQLite binaries, if someone pushes a

change to git and someone else pulls

that change, the user pulls the updated

JSONL file, merge conflicts are resolved

if there are any and the data is updated

in the SQLite database, meaning user A

will have the same issues as user B.

This is known as a two-way sync, which

is really [music] cool. But at this

point, we don't have any issues. So

let's create one which we can either do

in the CLI use the beads MCP server or

just ask your coding agent directly

which will create all these issues and

even put them under the epic with

correct P levels. From here we can

update or delete issues if we don't

think they're needed. Then when you're

ready you can get your agent to start

working on the issues and it will

automatically update them with the

relevant progress until there are no

more open issues and the project is

completed. Of course, from here you

could go on to add more issues, epics,

even use a project specific database.

And if you don't want to keep running

commands to view the status of issues,

you could use the web UI or even sync it

with Jira. And for large projects, if

your database ends up growing in size,

beads has a compaction feature that

reduces the size of a database by

compressing old closed issues while

preserving their essential context. But

why would I use beads over the more

popular specd driven development? Well,

the biggest benefit is context. With

spec kit and other tools, you have

markdown and prd files, some of which

the agent needs to load in entirely,

which takes up a lot of context. But

with beads, the agents can just query

the database to get the information it

needs, which works especially well for

issues with dependencies since the

relationship is more explicit instead of

being part of a list. Meaning the LLM

can just follow the dependency graph and

the human so me is never afraid of

clearing the context because information

is never lost. Not to mention the

two-way syncing mechanism is better for

multiple agents even on different

machines to collaborate on the same

project. Personally, the mental model of

having issues, assignees, estimations,

and epics is much easier for me to get

my head around instead of thinking of

PRDs. But I do prefer the emphasis on

planning when it comes to specd driven

development that seems to be missing

from beads. So, one approach would be to

create a detailed markdown file using

something like spec kit and then using

beads to create the epics and issues.

But maybe it's just overkill for a small

to medium project. Anyway, what are your

thoughts on beads? Are you planning to

use it in a future project? Let me know

your thoughts in the comments. Again,

don't forget to subscribe.