Deploy agents,
not infrastructure.

Push your code, connect your customers' tools, and let the platform handle browsers, approvals, credentials, and recovery.

The deployment platform for AI agents. Works with LangChain, CrewAI, OpenAI Agents, Playwright, Anthropic, and AutoGen.
Free tier · No credit card
app.runforge.sh / price-monitor / runs / run_7e834c
Platform
Projects
Approvals1
Integrations
Price Monitor
Overview
Runs
Customers
Settings
Customer: Acme Corp · 15.2s · 3 of 4 steps
Step timeline
open_page
webscraper.io · 2.1s
safe
extract_products
6 products · 1.8s
safe
sheets.append_row
Google Sheets · 0.4s
safe
email.send_report
via customer's Gmail
Approval required
Send price report to admin@acme.com
commit
Browser · Active
Lenovo IdeaPad
$321.94
Asus VivoBook
$356.49
HP EliteBook
$413.90
Dell Inspiron
$299.00

Trusted by developers
building production agents

<3m
Deploy time
0
OAuth code to write

Sandboxed execution

Isolated browsers per run

Encrypted credentials

AES-256 at rest, TLS in transit

Human in the loop

Approve before agents act

Crash recovery

Auto-resume from checkpoint

Everything you need to
deploy agents with confidence

Managed browsers

Isolated Chromium per run. Navigate, fill forms, extract data, screenshot. Sandboxed.

GmailConnected
SheetsConnected
SlackPending

Per-customer credentials

Customers connect their tools via OAuth. Platform encrypts and routes tokens per run.

open_page — 2.1s
extract — 1.8s
email.send — waiting
complete

Step observability

Full timeline for every run. Duration, browser state, pause points — all visible.

99.7% uptime

Crash recovery

Safe steps checkpoint. On failure, agent resumes from last checkpoint automatically.

Simple process,
powerful results

Go from agent code to production deployment in minutes, not weeks.

Get Started
01

Connect your repo

Point RunForge to your GitHub repository. We detect your agent framework, install dependencies, and configure the runtime.

02

Define your steps

Mark steps as safe (auto-run) or commit (require approval). The platform handles pausing, notifying, and resuming.

03

Ship to customers

Share a connection portal. Customers link their Gmail, Sheets, Slack. Your agent runs on their behalf — you focus on the product.

04

Monitor and iterate

Watch every run in real-time. See step timelines, browser screenshots, approval logs. Ship improvements with git push.

Approve from
anywhere

Get a Telegram message when your agent hits a commit step. See the context, tap approve, the run continues.

Approve or reject runs directly from Telegram
Trigger runs with /run price-monitor
View logs without opening the dashboard
RunForgeBot
online
Approval needed
price-monitor · email.send_report
Send report to admin@acme.com
Approve
Reject
2:14 PM
Approved
2:14 PM
Run succeeded
price-monitor · 4 steps · 8.3s
2:15 PM

Plans that scale
with your agents

Free
$0
Perfect for trying RunForge
Start Free
Included
  • 100 runs / month
  • 60 browser minutes
  • 1 project
  • Community support
Business
$99 /month
For teams with customer-facing agents
Request demo
Everything in Pro, plus:
  • 5,000 runs / month
  • 3,000 browser minutes
  • Unlimited projects
  • API product mode
  • Dedicated support

Your questions,
clearly answered

What is RunForge?

RunForge is a deployment platform for AI agents. You push your agent code, we handle the runtime — managed browsers, credential management, human approvals, crash recovery, and customer-facing connection portals.

What agent frameworks do you support?

Any Python-based agent framework — LangChain, CrewAI, OpenAI Agents, Playwright, Anthropic SDK, AutoGen, and more. If it runs in Python, it runs on RunForge.

How do customer credentials work?

Each customer gets a hosted connection portal where they OAuth-connect their tools (Gmail, Sheets, Slack). Tokens are AES-256 encrypted and automatically routed to the right run. You write zero auth code.

What are safe and commit steps?

Safe steps run automatically. Commit steps pause the agent and require human approval before proceeding — via the dashboard or Telegram. This gives you full control over what agents do on behalf of customers.

Is my data secure?

Yes. Browser sessions are sandboxed and isolated per run. Customer credentials are AES-256 encrypted. All data is transmitted over TLS. Sessions never leak between customers.

Can I use RunForge for free?

Yes. The free tier includes 100 runs per month, 60 browser minutes, and 1 project. No credit card required to start.

Gmail
Sheets
Slack
Notion
GitHub
SFSalesforce
16Calendar
Discord
Stripe
Jira
HubSpot
Zapier
WhatsApp
Twilio
Airtable
Telegram

Deploy your first agent.

Free to start. GitHub to running agent in under three minutes.