AI agents are a mess. Let's untangle them.
Build your AI agent in a friendly visual builder — no code, no prompt-engineering PhD. Then take it anywhere: export to Claude Code or Cursor, or deploy and run it live in the cloud.
you're a developer, not a format translator
Claude Code wants CLAUDE.md. Cursor wants .cursor/rules. Every platform wants something else entirely. You just want your agent to work.
# claude code wants this...
instructions: |
You are a helpful assistant...
# cursor wants this...
rules:
- description: "helpful assistant"
# and every other tool wants its own...
You are a helpful assistant that...
same instructions. a different file format for every tool. this is your life now.
Design once, visually
Point and click instead of writing config files
Export everywhere
Claude Code, Cursor — one click
No format headaches
phyla handles the translation for you
what's in the box
point and click beats copy and paste
Drag skills in, toggle behaviors, see your agent take shape in real time. No YAML. No mysterious config files. Just a builder that works the way your brain does.
build once, export everywhere
Write your agent config once and export to Claude Code or Cursor. Switch tools tomorrow without rebuilding from scratch.
or skip the files and run it live
Don't want to wire up config files? Deploy your agent straight to the cloud and run it live on Anthropic's Claude Managed Agents runtime — no servers, no setup, no API key to wrangle. You get monthly run credits and can top up anytime, then start and stream sessions from the app, API, SDK, or CLI.
steal skills from the community (legally)
Browse pre-built skills other developers have shared. Code review standards, commit message formats, framework conventions — grab what you need and make it yours. Why start from zero?
no account required. seriously.
Open the builder and start configuring. No signup wall, no 14-day trial, no “just enter your email” popup. Create an account later if you want to save your work or collaborate.
three steps. that's it.
design your agent
Open the builder and drag in skills, set behaviors, define what your agent knows. Takes about five minutes.
pick your platforms
Claude Code and Cursor — check the ones you use. Or just check all of them.
export and go — or deploy and run
Download the config files and drop them into your repo — or skip the files entirely and deploy your agent to the cloud, where it runs live in one click. Your call.
plays nice with your stack
who's building with phyla
from solo side projects to org-wide standards, phyla flexes to fit.
Developers
Configure AI assistants that know your stack, follow your conventions, and write code the way your team actually writes code. No more correcting the same mistakes.
Learn moreProduct managers
Build AI assistants that draft specs in your format, organize backlogs your way, and stop confusing “epic” with “feature.”
Learn moreTeams
Share agent configs across your org so the new hire's AI assistant works the same as the staff engineer's. Alignment without another Confluence page.
Learn morefrequently asked questions
Do I need an account to use phyla?
No. Open the builder and start creating — no signup, no email, nothing. You only need an account if you want to save your work or collaborate with teammates.
What platforms does phyla export to?
Claude Code (CLAUDE.md) and Cursor (.cursor/rules). More platforms are on the roadmap based on what the community asks for.
Is phyla free?
The builder is free with no limits on agents or exports. Paid plans exist for teams that need shared libraries, collaboration, and admin controls.
Can I use phyla with my team?
Yes. Team plans let you share agent configs, maintain a shared skill library, and make sure everyone's AI assistants follow the same standards.
What's a 'skill' in phyla?
A skill is a reusable block of instructions for a specific task — like writing commit messages in conventional commits format, reviewing PRs against your team's checklist, or following a particular framework's conventions. You can build your own or grab them from the community library.
Can I version-control my agent configs?
Yes. Exported configs are plain text files (Markdown, YAML, etc.) that live in your repo. Version them, diff them, review them in PRs — they're just files.
What happens to my data?
Your agent configurations are stored locally in your browser until you create an account. We don't train on your data, and we don't sell it. Paid team plans store configs server-side with encryption at rest.
How is this different from just writing CLAUDE.md by hand?
You can absolutely write config files by hand. phyla helps when you want to reuse skills across projects, export to multiple platforms at once, or share configurations with your team. Think of it as the difference between writing HTML by hand and using a design tool — both work, one is faster for most people.
Can I import my existing CLAUDE.md or .cursor/rules?
Not yet, but it's coming. We're building an importer that will parse your existing config files and bring them into the visual builder so you don't have to start over.
Is phyla open source?
The core builder is not open source, but the skill library is community-driven and all exported configs are plain text files you fully own. We're considering open-sourcing parts of the platform in the future.
your agent is only as good as its instructions
you've been writing config files by hand. there's a better way, and it takes five minutes to try. no account, no credit card.