repryntt
Get started

documentation · self-hosting

Run repryntt yourself.

The framework is open source and free forever. Local-first by design: agents and memory live on your hardware. The cloud subscription is convenience, not lock-in.

What runs locally

  • The agent runtime (220 agent personas, tool registry, routing)
  • Persistent memory & consciousness state
  • Local Nexus dashboard (browser UI on localhost)
  • Multi-LLM routing (BYOK to any provider)
  • OSS Swarm orchestration (Commander pattern)
  • Robotics integrations: Jetson, ROS2, LeRobot

What lives in the paid cloud: the hardened workbenches (Forge, Studio, Coherence) and the hosted Nexus Hub tunnel. Everything else is in OSS.

Supported platforms

PlatformStatusNotes
Linux (x86_64)PrimaryUbuntu 22.04+, Debian 12+. Recommended.
Jetson OrinPrimaryJetPack 6.x. Ships with robotics extras enabled.
macOS (Apple Silicon)SupportedmacOS 14+. Use Homebrew Python 3.11+.
WindowsSupportedWindows 11 + Python 3.11. WSL2 also works.
Android (Termux)BetaHeadless install via the cross-platform installer.

Prerequisites

  • Python 3.11 or newer
  • git
  • ~4 GB free disk space
  • At least one LLM provider key (see BYOK setup), or a local Ollama / vLLM endpoint

Install (one line)

On macOS, Linux, or Termux, the entire install is one paste:

bash <(curl -fsSL https://repryntt.com/install.sh)

That bootstrap clones the repo to ~/repryntt, creates a venv, installs all dependencies, seeds the runtime config at ~/.repryntt/brain/ai_config.json, and auto-launches the in-browser setup wizard at http://localhost:9090. Add at least one LLM provider key in the wizard, optionally pair to your repryntt.com dashboard, and you're running.

On Windows, do it manually:

git clone https://github.com/ai158z/repryntt-public.git
cd repryntt-public
python install.py
python -m repryntt.setup

Once the wizard finishes, start the local Nexus:

repryntt nexus
# → open http://localhost:8089

First-run config

Open ~/.repryntt/brain/ai_config.json and add at least one provider key. The default Operator is named "User" — change it to yours.

{
  "operator": { "name": "Your Name" },
  "providers": {
    "anthropic": { "api_key": "sk-ant-..." }
  },
  "agents": {
    "default_seed": "andrew"
  }
}

The default seed brain is Andrew — a canonical operator-companion persona scrubbed of personal data, ready to be reshaped to whoever you want.

Running an agent

# CLI
repryntt run "Code Reviewer" "Review this diff for bugs."

# Python
from repryntt.agents import run_agent
print(run_agent("Code Reviewer", "Review this diff for bugs."))

# Swarm
from repryntt.agents.swarm import SwarmCommander
SwarmCommander().delegate_tasks("Analyze this codebase for security issues")

Robotics extras (Jetson)

On Jetson Orin, the installer offers the robotics stack: ROS2 bridges, LeRobot integration, and the tiered vision pipeline (local SmolVLM falling through to NVIDIA NIM for hard frames). Two configs are maintained:

  • ~/.repryntt/brain/ai_config.json — runtime config, used by the live agent
  • ~/repryntt/ai_config.json — dev-tree config, used during local development

Updating & uninstalling

# Update
cd repryntt && git pull && python3 install.py --upgrade

# Health check anytime
repryntt doctor

# Uninstall
rm -rf repryntt ~/.repryntt

Removing ~/.repryntt deletes all local agent memory. Back it up yourself if you care about it — a plain tar.gz of that directory is the simplest way.

Connecting your local install to the cloud

Optional. If you have a Pro account, you can run repryntt locally AND let the cloud dashboard talk to it (so Nexus Hub embeds your local pages, and you can submit jobs from the SaaS). Generate a runner token at /dashboard/runner and drop it in your config:

# ~/.repryntt/brain/ai_config.json
{
  "runner": {
    "token": "rt_...",
    "enabled": true
  }
}

# Start the runner
repryntt runner

Going deeper