Welcome to RunLoop
Run a team of AI agents
Define goals, hire roles, set budgets. The control plane for autonomous software, in one self-hosted deployment.
npx agentai onboard --yes
pnpm dlx agentai onboard --yes
bunx agentai onboard --yes
Trusted by engineering teams
500+ companies use RunLoop
4.8/5
14-day free trial · Cancel anytime
What you get
Everything you need to ship autonomous products
Four primitives, one platform. No glue code, no half-built integrations, no surprises in production.
Orchestration
Coordinate teams of agents with delegation, dependencies, and shared context across runs.
Agent Runtime
Long-running agents with persistent state, resumable sessions, and graceful failure recovery.
Tracing
Every prompt, tool call, and decision logged. Replay any session, debug any failure.
Governance
Approval gates, hard budget caps, and audit logs. Autonomy without the runaway risk.
ABOUT US
Built For Builders, Not Buyers
At RunLoop, we believe great infrastructure disappears. Our job is to handle the boring parts of running agents so you can focus on what they're actually doing.
Learn More
What we offer
Agent Orchestration
Open Source Core
Real-Time Tracing
Budget Controls
Self-Hosted Deploy
$ npx agentai onboard --yes
→ Installing AgentAI runtime...
→ Setting up Postgres on port 5432...
→ Ready. Open http://localhost:3000
→ Installing AgentAI runtime...
→ Setting up Postgres on port 5432...
→ Ready. Open http://localhost:3000
pnpm dlx agentai onboard --yes
bunx agentai onboard --yes
BENEFITS
Production-Ready Out Of The Box
Skip the months of glue code. Ship autonomous agents with budgets, tracing, and governance built in from day one
Open Source Core
Self-hosted, MIT licensed, no vendor lock-in. Fork it, audit it, run it on your own infrastructure forever
Predictable Costs
Hard budget caps and soft warnings keep agents on a leash. No more $4,000 overnight bills from runaway loops
Full Observability
Every prompt, tool call, and decision logged. Replay any session, debug any failure, audit any run
Bring Your Own Model
Works with OpenAI, Anthropic, local models, custom endpoints. Switch providers without rewriting a single agent
8.4M+
Agent-Hours Run
Team Orchestration
Agents work in hierarchies with delegation, dependencies, and shared context — not isolated scripts running in parallel
Battle-Tested Patterns
Clone proven org configurations from the community. Marketing teams, support fleets, trading desks — ready to deploy
OUR PROCESS
From Zero To Autonomous In Three Steps
01
Install & Connect
Get the runtime running in under five minutes. One command installs everything you need, locally or on your own infrastructure.
npx install
Connect Models
Spin Up Workspace
02
Define Your Agents
Write your agents in ten lines of code. Set their goals, schedules, and budgets. Build an org chart with delegation and shared context.
Set Goals
Assign Roles
Define Budgets
03
Deploy & Monitor
Hit run. Watch agents work from one dashboard. Pause, override, or terminate any run. Every decision logged, every session replayable.
Approve & Launch
Live Monitoring
Replay Any Run
Use Cases
Run AI agents for Thousands of Jobs
One platform, every use case- Built for teams of agents
- Production-ready from day one
Testimonials
Marcus Thompson
Founder & CEO
The control plane for autonomous software
Imagine if internal tools actually worked. We've tried countless agent platforms and there is nothing quite like RunLoop.
Marcus Thompson
Founder & CEO
Built in the open, by the people who use it
It's rare to find a tool this powerful that's also actually open source. AgentAI is what every infrastructure company should aspire to be.
Marcus Thompson
Founder & CEO
Reliable enough to run unattended
We replaced three monitoring tools and a homemade cron setup with RunLoop. The budget caps alone paid for the migration in week one.
Marcus Thompson
Founder & CEO
The control plane for autonomous software
"We ship 5x the autonomous workflows we did six months ago. The tracing layer is the reason none of them break in production."
Marcus Thompson
Founder & CEO
The control plane for autonomous software
"We ship 5x the autonomous workflows we did six months ago. The tracing layer is the reason none of them break in production."
Marcus Thompson
Founder & CEO
The control plane for autonomous software
Imagine if internal tools actually worked. We've tried countless agent platforms and there is nothing quite like RunLoop.
Integrations
50+ integrations. Built-in.
Connect RunLoop to your model providers, runtimes, databases, and developer tools. If we don't have what you need, build it in 30 lines with our SDK.
Explore All Integrations
Problem Solved
What changes when you ship with RunLoop
Without RunLoop
Twenty terminal tabs open. No idea which agent is doing what. State lost on every restart.
With RunLoop
Tasks become tickets. Conversations stay threaded. Sessions persist across restarts.
Without RunLoop
One runaway loop burns through your entire monthly token budget overnight.
With RunLoop
Hard budget caps stop agents at the limit. Soft warnings at 80% give you time to react.
Without RunLoop
Recurring jobs you forget to schedule. Cron files scattered across machines.
With RunLoop
Heartbeats run scheduled work automatically. You supervise instead of executing.
Without RunLoop
One bug, lost in a chat scrollback. No idea what prompt or tool call caused it.
With RunLoop
Every prompt, tool call, and decision logged. Replay any session, debug in minutes.
Without RunLoop
Hand-rolled glue code that breaks every time a model API ships a new version.
With RunLoop
Bring-your-own-model abstraction. Switch providers without rewriting a single agent.
Pricing
Free forever to self-host. Pay when you scale.
Run RunLoop on your own infrastructure for free, or use our managed cloud when you need to focus on shipping
Enterprise
Organizations with compliance and scale needs
$89
/month
This plan includes:
- Single-tenant VPC deployment
- SAML, SSO, and SCIM provisioning
- Dedicated success engineer
- Custom SLAs and contracts
- SOC 2, GDPR, and DPA support
Foundation
Solo builders or self-hosted teams
$29
/month
This plan includes:
- MIT-licensed source code
- Up to 5 active agents
- Local Postgres or SQLite
- Community support on Discord
- Full feature parity with paid plans
Growth
Teams ready to ship without managing infra
$49
/month
This plan includes:
- Unlimited active agents
- Managed runtime & hosted Postgres
- 99.9% uptime SLA
- Email support, 24-hour response
- 90-day audit log retention
$ npx agentai onboard --yes
→ Installing AgentAI runtime...
→ Setting up Postgres on port 5432...
→ Ready. Open http://localhost:3000
→ Installing AgentAI runtime...
→ Setting up Postgres on port 5432...
→ Ready. Open http://localhost:3000
npx agentai onboard --yes
bunx agentai onboard --yes
Blog
Explore Product Latest Updates & Insights
.jpg)
.jpg)
FAQ
Everything else you might want to know
Quick answers to help you plan, design, and build with confidence
What is RunLoop and who is it for?
RunLoop is the open-source control plane for AI agents. It's built for engineering teams, AI startups, and anyone shipping autonomous software — from solo builders running 5 agents to enterprises managing fleets of 500.
How is RunLoop different from agent frameworks like LangChain or CrewAI?
Frameworks help you build agents. RunLoop helps you run them at scale. We sit one layer above — handling orchestration, budgets, observability, and governance for agents you've already built (in any framework).
Is RunLoop really open source?
Yes. MIT licensed, full source on GitHub, self-hosted forever. Fork it, audit it, run it on your own infrastructure with zero strings attached. The cloud version is paid, but the runtime is free.
Can I bring my own model?
Any model, any provider. OpenAI, Anthropic, local models, custom endpoints — plug it in and route tasks to it. Switch providers without rewriting a single agent.
Do I need to know how to code to use RunLoop?
For the OSS version, yes — you'll write agents in TypeScript or Python. The Cloud version has a visual builder for teams who want to configure agents without writing code.
Is there a free trial?
14 days on Cloud Pro, no credit card required. Hobby (self-hosted) is free forever — no trial needed, just install and run.
What happens when I hit a budget cap?
The agent auto-pauses. You get a soft warning at 80% of budget and a hard stop at 100%. Override anytime from the dashboard. No surprise bills.
Can I switch between Hobby and Cloud later?
Yes, anytime. Your data migrates automatically when you upgrade. Downgrade to Hobby and you keep self-hosting with the same setup.
Do you offer discounts for startups or non-profits?
50% off Cloud Pro for verified non-profits and academic institutions. Startups in YC, Techstars, or similar programs get 6 months free. Email sales for details.
What about token costs from model providers?
You bring your own model keys. We don't mark up token costs — bills come direct from OpenAI, Anthropic, or whoever you're using. AgentAI only charges for the orchestration layer.
How long does it take to get up and running?
Under five minutes for the OSS version. One command installs the runtime, sets up Postgres, and opens the dashboard. Your first agent runs in under an hour.
What support do I get on each plan?
Hobby gets community support on Discord (4,000+ builders, fast response). Pro gets email support with 24-hour response. Enterprise gets a dedicated success engineer and custom SLAs.
Can I migrate from another agent platform?
Yes. We have migration guides for LangChain, CrewAI, and homegrown setups. Most teams complete migration in under a week — your existing agent logic ports over without rewrites.
Do you offer onboarding help?
Pro plans get a free 30-minute kickoff call. Enterprise plans get a dedicated implementation engineer for the first 90 days. Hobby users get our Discord community and full docs.
What happens if something breaks in production?
Cloud Pro and Enterprise come with 99.9% uptime SLAs and a public status page. Critical issues get same-day response on Pro, dedicated incident response on Enterprise. Bugs in the OSS runtime are triaged daily on GitHub.
