Now available for Claude Code

Manage skills, rules,
and prompts for AI agents.
One source of truth.

The skill, rule, and prompt management platform for teams using AI coding agents. Centralize your coding standards in one place and deliver them automatically, every session, without burning tokens.

See it in action
No credit card requiredFree tier includes 2 projectsWorks with Claude Code, Cursor, and more
MCP-native skill and rule delivery
Works with Claude Code, Cursor, Windsurf, and more
Token-efficient lazy loading
Replaces scattered CLAUDE.md and Cursor rules files
The problem

Your AI coding agents keep making the same mistakes.

Your team has spent months building coding standards: how to write migrations, how to handle API versioning, how to structure tests. That knowledge lives in your heads, in scattered CLAUDE.md files, in .cursorrules that drift between repos, in code review comments that get forgotten.

Every Claude Code session starts from zero.The agent doesn't know your rules until something goes wrong. Then someone opens a PR that violates three conventions at once and the review becomes a lecture.

The alternative is stuffing everything into a single CLAUDE.md or AGENTS.md file, which costs thousands of tokens per session whether the agent needs it or not. You're paying for context that's irrelevant 90% of the time.

📂
Scattered CLAUDE.md and Cursor rules
Every repo has its own CLAUDE.md, AGENTS.md, or .cursorrules file. They drift, contradict each other, and nobody knows which one is canonical.
🪙
Token waste on every session
Loading all your coding standards upfront costs tokens whether they're relevant or not. You pay for context the agent ignores.
🔁
Same mistakes, every session
The coding agent doesn't know your team changed the migration pattern last month. It uses the old one. Again.
👤
No source of truth for agent context
Senior engineers review AI output and catch violations by memory. Your coding conventions aren't centralized, so every agent session is a coin flip.
How it works

From scattered files to one managed skill and rule library.

Replace scattered CLAUDE.md, AGENTS.md, and Cursor rules files with a centralized skill and rule management platform. Three steps, under four minutes.

1
Author your skills, rules, and prompts
Write them in plain markdown, or import your existing CLAUDE.md and AGENTS.md files. Organize by scope: org-wide, team, or project-specific. Set which ones load automatically and which are available on demand.
2
Connect via MCP
Run syphin init in any project directory. Syphin installs the MCP server and writes a config file. Claude Code, Cursor, and Windsurf pick it up automatically. No other setup required. Works in monorepos too.
3
Watch your agents improve
The coding agent searches your skill and rule library before starting work, loads only what's relevant, and follows your conventions, session after session. The Sessions tab shows exactly what was loaded and how it was used.
Live demo

From terminal to live session in 90 seconds.

This is what actually happens when a developer opens Claude Code in a Syphin-connected project.

Terminal · checkout-api
$ █
waiting for connection…
checkout-api · dev · jordan@acme.com
Features

Skill and rule management built for engineering teams.

More than a file host. A complete platform for managing AI agent skills, rules, prompts, and session telemetry.

Semantic skill search
Agents find relevant skills by describing what they're doing, not by guessing the right keyword. The same vector search powers both agent discovery and human authoring.
Token-efficient
Scope inheritance
Coding standards cascade from org to workspace to project. Define your Python conventions once at the org level. Override for specific teams or projects. No duplication across CLAUDE.md files.
Live session telemetry
See exactly what context every AI coding agent loaded, searched, and ignored. Session by session. Know whether your coding standards are actually being followed.
Local caching
Skill bodies are cached on the developer's machine after the first load. Published versions are immutable, so the cache never expires. Fast offline, fast always.
GitHub integration
Relevant skills surface as PR comments. CLAUDE.md and AGENTS.md files sync automatically. New repos get connected without anyone thinking about it.
Builder+Coming Soon
Skill A/B testing
Split-test description variants and measure which version gets loaded more often. Turn "does this rule work?" from a guess into a number.
Team+Coming Soon
Description quality linter
Skills only work if agents can find them. The linter scores your descriptions in real time and shows you what queries will surface each skill.
Review workflows
Draft, review, approve, publish. Skills go through the same deliberate process as your code, with inline comments and full version history.
Team+
Skills marketplace
Browse and import community coding standards for Prisma, Stripe, FastAPI, React, and more. Contribute your own. Syphin's quality bar means you can actually trust what you import.
Works with every MCP-compatible AI coding agent
Claude Code
Cursor
Cline
Windsurf
Any MCP agent

What teams are saying.

“We had 14 different CLAUDE.md files across our repos, each one a month out of date. Syphin replaced all of them with one place our whole team actually keeps current.”
SR
Sofia R.
Staff Engineer · Series B fintech
“The Sessions tab is what sold me. I can see exactly which skills Claude loaded for any given task. It turned a black box into something I can actually tune.”
DK
Dev K.
Engineering Lead · Developer tools startup
“I'm not a developer but I use Claude Code every day to build features. Syphin made it so the agent already knows how our team builds things. I don't have to explain it every session.”
MP
Maria P.
Product Manager · SaaS company
Ready to start

Your agents.
Your rules.
Every session.

Create a free account, centralize your skills, rules, and prompts, and start delivering them to every AI coding agent session.

View pricing