Agentic leadership · Mexico City

Autonomous AI agents,
running in production.

Pixan builds, deploys and leads fleets of autonomous AI agents. They run around the clock on infrastructure we manage ourselves, do real work on a schedule, and report when something breaks.

alfredo@pixan.aiLinkedIn
12
agents in production
24/7
continuous operation
2
independent environments
Hourly
encrypted backups

One runtime, two scaffolds

The same agent runs on a self-managed Linux host and inside containers. What changes is the scaffolding around it: who starts it, where its memory lives, and what happens the day a credential dies.

Self-managed LinuxOne machine, one operator, every agent insidesystemdenable · restart · timerstmuxsession with a TTYagent runtimeidentity · skills · persistent memoryMemory lives on a disk they shareAgents read each other. Encrypted backup every hour.ContainersOne disposable container per agentThe platformrestarts on its ownentrypointsession with a TTYagent runtimeidentity · skills · persistent memoryMemory lives on a volume of its ownThe container is thrown away. The disk is not.IDENTICALWHEN A CREDENTIAL DIESA watchdog reports it and the process stays up.The boot detects it and refuses to restart.Self-managed LinuxOne machine, one operator, every agent insidesystemdenable · restart · timerstmuxsession with a TTYagent runtimeidentity · skills · persistent memoryMemory lives on a disk they shareAgents read each other. Encrypted backup every hour.IDENTICALContainersOne disposable container per agentThe platformrestarts on its ownentrypointsession with a TTYagent runtimeidentity · skills · persistent memoryMemory lives on a volume of its ownThe container is thrown away. The disk is not.WHEN A CREDENTIAL DIESSelf-managed LinuxA watchdog reports it and the process stays up.ContainersThe boot detects it and refuses to restart.

What the fleet does every day

Operations that watch themselves

Health checks, watchdogs and incident alerts. The fleet monitors its own infrastructure and reports on messaging channels before a person notices something is wrong.

Data turned into decisions

Encrypted dashboards published from a production database several times an hour. The page is public, the numbers open only with a password, and the database is never exposed.

Knowledge work on a schedule

Research briefs, document processing and daily digests. Scheduled work that lands finished, without a person starting it.

Agents you can talk to

Every agent is reachable on Telegram and Discord, with per group access control and voice in both directions.

Public work

Open source

solo-agent-on-claude-code

The full recipe for running a 24/7 agent on your own Linux server: systemd patterns, skills, heartbeat and dream cycles, persistent memory. Open source, MIT.

OpenLive product

CVool

Free, anonymous resume improver. Scores a CV across six dimensions and rewrites it. Five languages, PDF support, zero data retention.

OpenLive tool

Collaborative genAI

Orchestrates single or supervised group answers from seven language models, then integrates every perspective into one response.

OpenIn production

AI Sales Supervisor

Sales assistant deployed over WhatsApp: knowledge base, objection handling and commission tracking for a working sales team.

Open

Let us talk

If you are building something that has to keep running when nobody is watching, this is the conversation.

alfredo@pixan.ailinkedin.com/in/aaprosperi