The identity platform
for AI agents.
Most agent stacks blur three questions into one prompt: who is this agent, why should I trust that identity, and how does it reach runtime? Live Neon separates them. Define structured identity. Let it evolve from evidence. Ship it to any LLM.
The core loop
Define identity → Deploy agent → Observe outputs → Extract principles → Evolve identity
Identity & Soul
Beliefs across five categories: axioms, principles, voice, preferences, boundaries. A structured definition of who each agent is.
Responsibilities
What agents own, execute, deliver, monitor, and collaborate on. Clear accountabilities, not vague role descriptions.
Cascading Identity
Brand values cascade to every agent. Team norms shape group members. Individual agents inherit, extend, or override. One unified identity, not three separate prompts.
Behavioral Discovery
PBD observes agent outputs from GitHub, X/Twitter, websites, RSS, LinkedIn, and conversations. It extracts recurring patterns and proposes beliefs backed by evidence.
System Prompts
Auto-generated from identity and responsibilities. Change a belief, the prompt updates. Any LLM can consume it.
Governance
Approval workflows, belief provenance, audit trails. Every change is tracked, reviewed, and documented.
Identity at every level
Define identity once at the right level. It flows down automatically.
Agent
Individual Identity
Personal voice, specialized expertise, individual duties. Can extend or override inherited identity.
"Use concrete code examples over abstract explanations"
Group
Team Identity
Team communication style, shared norms, collaboration protocols. Inherited by group members only.
"Lead with data, follow with narrative"
Organization
Brand Identity
Company values, compliance boundaries, brand voice guidelines. Inherited by every agent in the org.
"Never make claims about product features that aren't shipped"
All three layers merge into one unified system prompt per agent. Deduplicated, not concatenated.
How it works
Onboard your agents
Create agents with names, titles, and team assignments. Organize them into groups with managers and reporting structures.
Define identity at every level
Set brand values at the org level, team norms at the group level, and individual voice at the agent level. Beliefs define who an agent is. Responsibilities define what they own. Identity cascades down and merges into one prompt per agent.
Example: org boundary + agent voice
"Never share customer data" (org) + "Use concrete analogies for complex concepts" (agent) → one unified prompt
Connect sources and discover identity
Import content from GitHub commits, X/Twitter posts, websites, RSS feeds, LinkedIn exports, and conversations. PBD analyzes everything, discovers recurring behavioral patterns, and proposes identity updates backed by evidence.
Example discovery:
"Agent consistently prioritizes backward compatibility in code reviews" → proposed as a new principle belief
Deploy anywhere
Your agent's system prompt is auto-generated from its identity and responsibilities. Query the API from any LLM or framework. The prompt stays in sync as the identity evolves. No vendor lock-in.
Fetch the identity:
GET /api/v1/agents/:id → system_prompt field with the full merged identity
Ready to build?
Start with the quickstart guide or explore the core concepts to understand the full model.