Free · open source · runs on your hardware

Install repryntt.
Run the full autonomous
AI system yourself.

220 specialists. Persistent memory. Multi-agent orchestration. Your device, your keys, your data. Free forever — and when you want the team that runs around it, membership is $29/mo.

install in one line — macOS · Linux · Termux

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

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

Source on GitHub →Become a member — $29/mo

MIT · BYOK · ~$1–10/hr at 40 RPM on frontier models · yes the Y is on purpose

220

specialist agents

380

built-in tools

workbenches

$29

membership

Not a chat.

220 specialists with tuned personas, tool policies, and a verify-before-deliver loop. Colleagues, not autocompletes.

Persistent.

Memory survives restarts. Files on your disk. Agents pick up where they left off.

Local-first.

The OSS runs entirely on your hardware. Air-gappable. BYOK across 6 providers.

Robotics-ready.

Same agent brain runs Jetson + ROS2 + LeRobot. Software and physical machines from one runtime.

Full feature breakdown →

the membership

Membership isn't a paywall.
It's joining.

The OSS gives you the engine. Membership gives you the team that runs around it. Hosted workbenches we build for ourselves. AI staff we put in charge of our own ops. Inner-circle access while we figure out what autonomous business actually looks like.

what members get today

Already shipped.

★6 hosted workbenches — Forge, Studio, Coherence, Swarm, Nexus Hub, Agent Studio.
★Universal inbox + Slack — Run agents from any channel. Slash command in; results message back.
★Brain Pulse — Live activity dashboard. See agents working in real time.
★REPRYNTT_API_KEY — Drop agents into your CI, scripts, automation.

coming, members first

The agents we're building to run ourselves.

We're putting our own agents in charge of how we run repryntt. Members get the front-row seat — and benefit from every workflow we ship.

→Member intelligence briefs — Weekly research drops from our scientific agent team.
→AI-run support — Agents triage your tickets first; humans take over when it matters.
→Monthly agent drops — New specialists added every month. Members first.
→Roadmap vote — Members pick what ships next.

why now

The first 100 members lock in the floor.

$29/mo, first dibs on every member-only drop, and influence on what we build next. Once we hit 100 the floor moves. Get in while it's flat.

Become a member — $29/mo See the full pitch

7-day free trial · cancel one-click · BYOK = no token markup

honest comparison

repryntt vs the peer projects.

We're in the same neighborhood as openclaw (the agent framework whose persistent-files method we borrowed for Brain Builder) and Hermes Agent (NousResearch's function-calling-tuned model line). Each is best at something different. Here's the honest table.

Capability	repryntt	openclaw	Hermes Agent
Specialist roster shipped out of the box	220 across 33 departments	0 — you build your own	0 — model + tools, no roster
Persistent memory across restarts	Yes — file-backed openclaw method	Yes — invented this pattern	Session-scoped via tool-state
Built-in tool registry	380 tools, 35+ categories, growing	Small core — bring your own	Grows with model fine-tunes
Multi-agent orchestration	Swarm Mode + Forge multi-step pipeline	Possible, not first-class	Possible, not first-class
Code generation with real test execution	Coding Forge — Architect → Tester	No built-in pipeline	Strong function-calling, no harness
Robotics / embodied agents	Jetson + ROS2 + LeRobot in OSS	Software-only	Software-only
Function-call model quality	BYOK — uses upstream models as-is	BYOK — uses upstream models as-is	Custom-tuned Hermes line excels here
Hosted dashboard option	$29/mo membership	Self-host only	Self-host only
Framework leanness	Opinionated, batteries-included	Lean — pure framework, your decisions	Model-first, slim runtime
License	MIT	Apache-2.0	Mixed — model licenses vary

Where openclaw wins: if you want a pure framework, no opinions, no roster. Build everything yourself, exactly the way you want it.

Where Hermes wins: if your bottleneck is function-call accuracy on a specific fine-tuned model. The Hermes line is purpose-built for that.

Where repryntt wins: if you want a complete system today — 220 specialists, 380 tools, six workbenches, robotics, and a dashboard — without building the harness yourself.

developer-first

Drop into your code.

Generate a REPRYNTT_API_KEY. Bearer token from any language, any framework, any CI.

Read the quickstart →

curl

$ curl https://api.repryntt.com/v1/agents/run \
   -H "Authorization: Bearer $REPRYNTT_API_KEY" \
   -d '{
     "agent": "Code Reviewer",
     "task": "Review this PR for bugs"
   }'

{
  "job_id": "job_a1b2c3",
  "agent": "Code Reviewer",
  "output": "Reviewed. Two issues found: ...",
  "model_used": "claude-sonnet-4",
  "tokens": { "prompt": 412, "completion": 89 }
}

Stop renting AI.
Run it yourself.

One paste, ten minutes, and you've got the full autonomous system on your device.