Beads vs. Google Antigravity

[rjbudzynski]

Basically what the title says.
Google Antigravity seems to have its own system for tracking work and managing context. AFAICS it misses Beads' git integration and is therefore suitable only for a single user/developer scenario. After playing around for a bit I still haven't decided whether Beads can supplement Antigravity's own system, or should I persuade the agent to use Beads only (looks hard to achieve), or maybe just not use Beads with Antigravity at all?
I'd appreciate any insights.

[rbergman]

I have experimented with this and it's viable. I would say it "works most of the time" to "persuade the agent to use Beads only." I believe I just used beads' OOTB onboarding guidance (but also had the MCP installed, which may or may not have helped). It still wrote its own plans and occasionally tasks file, but when I caught it doing that I would tell it to convert it to beads and toss it. It's imperfect and I could see its thinking arguing with itself over it, but it was mostly going with it. This was a couple weeks ago now, so YMMV -- I stopped using Antigravity for now due to usage limits.

[anntnzrb]

To make agents prefer Beads over Antigravity:

1. Onboard explicitly: bd onboard

2. Add to AGENTS.md/CLAUDE.md:

BEFORE ANYTHING ELSE: 
- Check for Beads: bd ready
- Use "bd create" for ALL new tasks
- Sync before finishing: bd sync

3. Create session-start hook (.claude/hooks/session-start.sh) to auto-init Beads.

The agent will initially argue with itself - this is normal. After 2-3 sessions with consistent redirection, it learns. Beads git integration is the key differentiator for multi-agent workflows.