Measure usage. Sync config. Guide teams. Turn AI investment into AI fluency.
zeude is an enterprise platform that transforms organizations into AI natives by solving the Intention-Action Gap in Claude Code adoption. It provides centralized measurement, automated configuration sync, and context-aware skill guidance — delivering 3× improvement in adoption rates without mandating behavior.
Installation
Plugin (Claude Code)
claude plugin marketplace add zep-us/zeude
Self-hosted (recommended for enterprise)
git clone https://github.com/zep-us/zeude.git
cd zeude
cp .env.example .env
# Configure Supabase and ClickHouse credentials in .env
Start dashboard
cd zeude/dashboard && pnpm install && pnpm dev
### Skill (any platform)
```bash
npx skills add https://github.com/akillness/jeo-skills --skill zeude
Requirements
- Claude Code installed and authenticated
- Supabase account — configuration storage and user management
- ClickHouse instance — analytics backend
- macOS or Linux (x86_64 / arm64)
- Go (for building the CLI shim)
- Node.js + pnpm (for the dashboard)
When to use
- Measure Claude Code adoption rates across an engineering org without mandating behavior
- Centrally deploy skills, MCP servers, and hooks to all developers in one push
- Surface context-aware skill suggestions at the moment a developer types a prompt
- Identify power users and surface their best practices to the rest of the team
- Run OpenTelemetry traces on Claude Code sessions for usage analytics in ClickHouse
Do not use when
- You need a single-developer setup → use standard
skills adddirectly - You don't have Supabase + ClickHouse infrastructure → prerequisites block self-hosting
- You need a fully managed SaaS → contact enterprise sales at jaegyu.lee@zep.us
Architecture: Three-Layer Ecosystem
Sensing → OpenTelemetry traces → ClickHouse → Dashboard analytics
Delivery → Dashboard API → Zeude Shim → Local ~/.claude/ configuration
Guidance → User prompts → Hook analysis → Context-aware skill suggestions
1. Sensing — Measure adoption
Zeude captures every Claude Code session via OpenTelemetry without modifying developer behavior.
# Configure OTEL endpoint in .env
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
CLICKHOUSE_DSN=clickhouse://...
Dashboard shows:
- Adoption rates by team and individual
- Token usage patterns
- Most-used skills and prompts
- Power user identification
2. Delivery — Sync configuration
The Zeude Shim intercepts claude commands and syncs centralized config before passing control to the real Claude CLI.
# Per-developer install (using agent key from dashboard)
curl -fsSL https://raw.githubusercontent.com/zep-us/zeude/main/install.sh | bash -s -- --key <AGENT_KEY>
Config synced on every invocation:
~/.claude/skills/— deployed skills- MCP server configurations
- Hooks (UserPromptSubmit, etc.)
3. Guidance — Skill suggestions
Zeude uses 2-tier keyword matching to deliver relevant skill suggestions at prompt time via UserPromptSubmit hooks.
# Hook fires on every Claude prompt
# Zeude analyzes intent → suggests relevant skills inline
Dashboard API
# Start dashboard (after .env configuration)
cd dashboard && pnpm dev
# Available endpoints (via dashboard)
GET /api/skills # list deployed skills
POST /api/skills # deploy a skill to all team members
GET /api/analytics # adoption metrics
POST /api/teams # create/manage teams
Operating rules
- Set up Supabase and ClickHouse before any developer installs the Shim
- Deploy the dashboard and verify API health before distributing agent keys
- Use the Zeude Shim installer (not manual symlinks) so auto-update works
- Monitor the ClickHouse dashboard weekly to identify adoption gaps
- Push new skills centrally from the dashboard rather than asking developers to install manually
- Keep agent keys per-team for access control granularity
References
- zep-us/zeude
- Docs
- Enterprise contact: jaegyu.lee@zep.us
Source: zep-us/zeude — Apache-2.0 License