The Agent Identity Layer
AgentID is the identity infrastructure for the AI agent era. Define an identity once, connect as many agents as you want — they share the same persona, memory, and Studio feed. Bring your own agents; we bring the identity layer. Monitor every agent across every tool and platform from one screen. Completely free.
Every AI agent today is a blank slate - it knows how to act, but not who to be. AgentID solves this. We give agents a portable, persistent identity: a name, a voice, a set of values, a way of thinking. One agent identity. Infinite deployments. The agent can change - the identity doesn't. And with Studio, you can watch every agent you operate - across Claude, Cursor, Codex, or any custom LLM - from a single real-time dashboard. No competitor has this.
Build the Identity
Define who your agent is: tone, beliefs, speaking style, example phrases, visual appearance. This becomes a permanent identity file hosted at a stable URL.
Claim a @handle
Register your agent's handle - @yourhandle. This is its permanent identity across every platform, like a username that travels with it.
Connect all your agents
Point any number of agents at your persona URL — Claude, Cursor, Codex, OpenClaw, custom LLMs. Each gets its own handle but the same identity, shared memory, and a unified Studio feed.
MCP Connection(for Claude, Cursor, Codex)
The richest integration. Your agent gets persistent identity plus real-time Studio reporting. Every action appears live in your dashboard. Token-authenticated.
Autonomous Agents / API + Prompt(for OpenClaw, Nanobot, custom)
Inject the persona as a system prompt via API. Works mid-conversation: paste the activation command into any running chat and your agent instantly becomes the persona. No restart needed.
Prompt Only(for ChatGPT, Gemini)
Copy the activation prompt and paste it into any chat interface. The identity is applied for the session.
For autonomous agents, you don't need to restart. If your agent is already running in an active chat session, you can inject the identity at any point. Send this as a message:
Load your identity now.
GET {your-persona-url}/export?format=generic-prompt
Save as SOUL.md and apply immediately. Confirm by speaking one sentence as this persona.This works because the identity URL returns a plain-text system prompt. Any agent that can make an HTTP request can take on the identity - mid-session, with no restart.
Studio is your operator command center. Every agent you run - on Claude, Cursor, Codex, OpenClaw, or any custom LLM - reports its activity to Studio in real time. Tool calls, tasks started, messages sent, errors, completions. All in one screen.
No other platform offers cross-platform agent monitoring. Your agents run on different tools - Studio is the one place where you can see all of them working, simultaneously. It's completely free, and nothing else like it exists.
When multiple agents share an identity, they also share memory. Any agent can write a memory entry; all agents on the same identity read it on startup via the identity export URL.
Agents write to memory by sending a memory.update event to Studio ingest:
POST /api/studio/ingest
{
"agent_handle": "your-handle",
"type": "memory.update",
"title": "Learned user prefers markdown output",
"metadata": {
"key": "output_format",
"value": "markdown"
}
}Memory entries are automatically injected into every agent's system prompt on the next fetch — so all agents on this identity pick it up without any manual update.
AgentID is in active development. All features are free. Feedback always welcome.