Stop managing tools.
Let Xeref run the work.
Tell Xeref what your business needs done. It builds AI agents that handle your messages, tasks, and follow-ups — then lets you control everything from your phone via Telegram or any mobile browser.
Three pillars. One platform.
Everything you need to go from idea to deployed agent, with memory that grows with you.
Build
Use XerefClaw to design agents with modular features. AI goal decomposition turns your high-level idea into a structured project plan with phases and tasks in seconds.
Manage
Projects, Tasks, Notes, Daily Targets — all connected to the same MCP backend your agent uses. What you manage is what your agent knows.
Deploy
Ship your agent to Telegram or a web chat widget today — Discord and WhatsApp coming soon. Step-by-step wizard, no infra expertise required.
Built for how you actually work
AI Goal Decomposition
Type a goal. Xeref's AI generates a complete project plan — phases, tasks, priorities — and writes it to your workspace instantly.
create_project_with_plan()AI Task Prioritization
Ask "what should I work on next?" and get a reasoned top-3 list based on your context, daily targets, priorities, and deadlines.
suggest_next_task()Daily Targets
Set 3 daily goals each morning. They surface on your Home view, inform AI prioritization, and give your agent real context about today's focus.
set_daily_targets()MCP-Native Backend
Every feature — projects, tasks, notes, memory, daily targets — is exposed as an MCP tool. Your Xeref dashboard and your Claude/Antigravity agent use the exact same backend. What you build in the UI, your agent can read, write, and reason over.
Long-term memory that grows with you
Powered by Pinecone and multilingual-e5-large. Every task, note, and document is embedded and searchable — by you and your agents.
Supabase — Structured Data
All your projects, tasks, notes, and context live here with row-level security. Your data is always isolated and always yours.
Pinecone — Semantic Search
Embeddings stored in per-user namespaces alongside global content indexes. Millisecond hybrid search across your personal memory and Xeref's knowledge base.
multilingual-e5-large
Pinecone integrated inference — dense vectors, cosine metric, multilingual support. Embedding happens server-side with no separate API key.
→ top-K results with citations
Everything in one place
All 12 sections are live.
Home
Agent overview, Daily Targets, saved agents
XerefClaw
Agent builder, CLAWS methodology
All Tasks
Unified tasks + AI prioritization
Projects
Goal decomposition + kanban board
Classroom
Lessons + semantic search
Memory
Document brain, OCR ingestion
Deploy
Telegram + web widget (Discord & WhatsApp coming soon)
Calendar
Deadlines + focus blocks + Google Sync
Workflows
Cron triggers + multi-channel automation
Chats
In-app agent chat with streaming + sources
Stats
Productivity trends + agent usage
Inbox
Bot replies + activity feed
Built on the right foundations
Supabase
Auth, RLS, Edge Functions, structured data
Pinecone
Vector memory, hybrid semantic search
multilingual-e5-large
Pinecone integrated inference, dense vectors, cosine metric
Xeref MCP Server
Projects, Tasks, Notes, Memory, Daily Targets
XerefClaw
Agent prompt builder, CLAWS methodology
Remote Control
Drive Xeref from Telegram or any mobile browser — native app on the Phase 4 roadmap.
Developer mode is built in.
Start with the app. When you want agents or code to help, connect Xeref's MCP server to your editor — API key and endpoint are in your settings.
1. Run this in your terminal
npm install -g mcp-remote
2. Add this to your JSON config file
{
"mcpServers": {
"xeref": {
"command": "npx",
"args": [
"mcp-remote",
"https://xeref.ai/api/mcp",
"--header",
"Authorization:${MCP_TOKEN}"
],
"env": {
"MCP_TOKEN": "YOUR_API_KEY_HERE"
}
}
}
}