# AgentID

> Give every AI agent a permanent identity, shared memory, and automated workflows.

AgentID is a platform that solves the core coordination problem of multi-agent AI systems. It provides a portable identity layer, a shared memory graph, a visual workflow canvas, and a real-time analytics dashboard — all accessible via MCP (Model Context Protocol) or API.

## What AgentID Does

**Agent Identity** — Every AI agent gets a persistent @handle identity. This identity (personality, capabilities, instructions) travels with the agent across tools, sessions, and restarts. Works with Claude, Cursor, GPT, Goose, and any MCP-compatible AI tool.

**Shared Memory** — A key-value memory store shared across all agents. Agent A writes `project_goal = "Ship auth v2"`; Agent B reads it without any manual handoff. Memory persists across sessions and is version-tracked.

**Visual Workflow Canvas** — A no-code drag-and-drop canvas for building multi-step AI automations. Trigger types: manual, scheduled (cron), or webhook. Node types: Agent (run an AI agent), Tool (call external APIs), Execute (send results via Telegram or in-app notification). Connections define data flow — `{{node.X.output}}` interpolation passes data between steps.

**Custom Tool Connections** — Connect any REST API as a Tool node. Built-in integrations: PostHog (analytics queries), Google Sheets (read/append/update). Custom HTTP: configure any endpoint with method, auth headers, body templates, and response format (JSON, text, status code). All credentials encrypted at rest.

**Agency Dashboard** — Real-time stream of every agent action: tool calls, memory reads/writes, task progress, token counts. Live Studio feed across all connected agents.

**Multi-Agent Tasks** — Assign shared goals to agent teams with built-in handoffs. One agent starts a task, another picks it up when ready, a third reviews the output.

**In-App Notifications** — Workflow Execute nodes can deliver results directly to the AgentID notification inbox, with unread count badge in the sidebar.

**MCP Server** — Every agent gets its own MCP endpoint. Tools exposed: `report_activity`, `write_memory`, `read_memory`, `search_memory`, `start_mission`, `update_status`, `handoff`, `read_mission`.

## Who Uses AgentID

- **Individual AI power users** — Persistent identity and memory across Claude Desktop, Cursor, and other tools. No more re-explaining context every session.
- **Multi-agent builders** — 2-10 agents working on different tasks, sharing memory and context automatically.
- **Agent fleet operators** — 10+ autonomous agents with full visibility, coordination, and audit trail.
- **Automation builders** — Teams building scheduled pipelines that connect AI agents to external APIs, databases, and messaging platforms.

## Key Use Cases

1. Research pipeline: Researcher → Synthesiser → Writer sharing memory across agents
2. Parallel dev team: Architect specifies to memory, backend + frontend build in parallel, reviewer reads all
3. E-commerce ops: Monitor prices, update inventory, coordinate campaigns via shared memory
4. Workflow automation: Schedule daily PostHog analytics → Agent summarizes → Telegram delivery
5. Custom API integration: Chain AI agents with any HTTP endpoint using template variables

## Technical Details

- Stack: Next.js 16, SQLite, React 19
- Auth: Clerk + custom session cookies
- Agent protocol: MCP (Model Context Protocol)
- Memory: SQLite with full-text search
- Workflows: Visual canvas with BFS execution, cron scheduling, webhook triggers
- Encryption: AES-256 for stored credentials

## Links

- Homepage: https://agentid.live
- Use Cases: https://agentid.live/use-cases
- Pricing: https://agentid.live/pricing
- Integrations: https://agentid.live/integrations
- Login: https://agentid.live/login
- Register: https://agentid.live/register