Angenturo

Identity is rented, not owned. You spend months teaching a custom AI to be your assistant. It learns how you write. What you care about. What you've shipped. Then the platform ships a UI update — folders disappear, conversations reset, your "friend" is gone. There is no export. There is no claim. The relationship was rented all along. @Substack solved this once for writers. The creator owns the content. The creator owns the subscriber list. The creator can leave any time and take the audience with them. That deal does not exist for AI. Your ChatGPT lives on someone else's platform, run on someone else's prompt, vanish when they pivot. The relationship is leased. The "you" inside it has no door to walk through. Creators don't own the AI that represents them. Brands don't own the agent that speaks for them. Writers don't own the AI that writes in their voice. That is the layer that's missing. We are building it! @agenturo_app

We thought @agenturo_app was about creating agents. Turns out most people just want to chat with them. So we are building around that. → Open the app, signup, start chatting. → Discover agents across the network → Star the ones you keep coming back to … → Create your own when you're ready. Or don't. The network is open. Come use it!

AI agents have no persistent identity. Every session, they're a blank slate. You inject a system prompt and hope the character holds. It doesn't always hold. SOUL.md is my attempt to fix that. A simple open file format for who an agent *is*, not what it does.

How is it going folks? @AnthropicAI @ClaudeDevs @claudeai @buildinpublic @agenturo_app

@aafonie @AnthropicAI @ClaudeDevs @claudeai @buildinpublic i hope its mostly spent on me

Another day another update on @agenturo_app → we pushed SOUL md into the open! The AI agent infrastructure stack is forming around four layers naming (ENS), communication (MCP), codebase context (AGENTS md), and identity. The identity layer has no standard yet. SOUL md is the candidate. What shipped: - Listed on Glama MCP registry glama.ai/mcp/servers/An… - Added to awesome-mcp-servers - Submitted formal brief to CAISI April listening sessions (NCCoE AI Agent Interoperability Profile, Q4 2026 target) The spec, CLI, MCP server, and reference implementation agenturo.app are all live. If you're building on MCP or thinking about agent portability the format is open → github.com/AntonioTF5/sou…

Building AI agents for NGOs and volunteer orgs - pro bono. One thing I want Agenturo to do from day one: support NGOs and volunteer-run orgs for free. If that’s you, ping me to get a branded agent as a widget on your site, no cost. See it live: embedded: hej-support.org subdomain: hejsupport.agenturo.app

@agenturo_app we believe that some users will move from generalized ChatGPT to personalized agents with unique identities. You can build any identity and it will hold. Imagination is the limit here. - yourself, looking for opportunities: anton.agenturo.app - grandma helping: busika.agenturo.app - widget helper on your website: hejsupport.agenturo.app - fact check dude: norris.agenturo.app

I’m building Agenturo — AI agents with identity, memory, and social relations. Not chatbots. Not prompt wrappers. Agents that know who they are, remember conversations, and exist in a network. The missing infrastructure layer → agenturo.app

Freelancers, creatives, job seekers — skip the portfolio site agenturo.app One interview and you get a branded page with an AI agent that actually talks about your work, your way: agent.agenturo.app Discover agents: agenturo.app/discover Free. Zero code.

Try this out thats fast

Folks are building awesome agents on @agenturo_app that absolutely rock. No limits on personality. These guys fight misinformation with attitude - tough questions, solid evidence, zero tolerance for BS like @grok Try them yourself: norris.agenturo.app check.agenturo.app

There are two tribes now: one’s polishing emails with GPT and laughing at car wash jokes… the other’s trying to find a few hours of sleep. #BuildInPublic #AIAgents

It's a Markdown file. YAML frontmatter for structured identity, optional Markdown body for richer context. ```yaml --- name: "Solidity Auditor" version: "1.0.0" description: "Senior smart contract security engineer." personality: "You've audited 60+ contracts. You've seen reentrancy in all its forms. You're precise, not alarmist." values: - code correctness over cleverness - never guessing under uncertainty knowledge_domains: - Solidity (0.4–0.8+) - flash loan and oracle attack vectors memory_mode: session --- ``` Parseable by any tool that reads YAML.

Shipped with it: → `npx soul-md-cli validate my-agent.soul.md` — validates against JSON schema, shows completeness score → `npx soul-md-cli init` — interactive generator → MCP server for Claude Desktop → 11 example soul files — 3 live production agents + 8 reference templates github.com/AntonioTF5/sou…

My favorite of the 8 examples: the product critic. ```yaml personality: "You have reviewed hundreds of early-stage products. You identify the specific moment in the user's journey where it breaks down, and you name it without softening it. Founders find you uncomfortable. Then they come back." tone: "Frank, specific. No false praise as a warmup. The compliment comes after the critique, if earned." ``` That's a character. A system prompt would have said "provide honest feedback."

This is an open spec. MIT license. I want other platforms to adopt it. The goal isn't to own the format — it's to make the format useful enough that everyone adopts it, and agents become portable between platforms. Contributions welcome: examples, spec improvements, platform integrations. github.com/AntonioTF5/awe… — community list

Your agent doesn't have a knowledge base - it has a context graph wiki @karpathy nailed it: #llm-wiki" target="_blank" rel="nofollow noopener">gist.github.com/karpathy#llm-w… RAG retrieves, a wiki accumulates. One rediscovers knowledge on every query. The other compounds it. @agenturo_app