Runloop

Code executionMIT

Runloop is an infrastructure platform for running AI agents on isolated cloud sandboxes called Devboxes. A developer or agent creates a Devbox, writes files into it and runs commands through the Python or TypeScript SDK, the command-line tool or the API. The home page also describes SSH and IDE access, interactive shells, websockets and Docker support inside Devboxes. Enterprise plans can deploy Runloop into the customer's own AWS, GCP or Azure account. Pricing combines a free Basic plan and a Pro plan with separately billed usage. The site says it serves AI-first teams that are building coding agents.

Pinout

  • MCP server: not checked yet
  • HTTP API: supported, verified
  • command line: supported, verified
  • Python SDK: supported, verified
  • TypeScript SDK: supported, verified
  • OpenAPI description: not checked yet
Where each pin comes from
InterfaceStatusSource we readRead on
MCP serverNot checked yet——
HTTP APISupported, verifiedHome page FAQ: API access with link to API reference
Command lineSupported, verifiedHome page, Get Started: CLI Tool
Python SDKSupported, verifiedGitHub api-client-python: pip install runloop_api_client
TypeScript SDKSupported, verifiedHome page, Get Started: TypeScript SDK for Node.js
OpenAPI descriptionNot checked yet——

Facts

Facts about Runloop
LicenseMIT (Python SDK repository). GitHub runloopai/api-client-python: MIT license, read .
Platformscloud; Runloop infrastructure can be deployed into your own cloud account (Enterprise VPC deployment). Home page: Deploy to VPC in an existing AWS, GCP or Azure account; pricing page lists Deploy to VPC under Enterprise, read .
PricingBasic: free plus usage; Pro: $250 per month plus usage; Devbox CPU $0.108 per CPU-hour. Pricing page: plans and usage rates, read .
Latest SDK releasev3.0.0 (Python SDK, dated 2026-10-23). GitHub api-client-python releases, Latest badge, read .
Python requirementPython 3.9 or higher (SDK). GitHub api-client-python README, read .
TrialPro trial with $50 usage credit, no credit card required. Pricing page, free trial section, read .
Last commit. GitHub API, latest commit on the default branch of runloopai/api-client-python, read .

Call example

from runloop_api_client import RunloopSDK

runloop = RunloopSDK()  # Uses RUNLOOP_API_KEY environment variable by default

# Create a sandbox, run a command in it, and clean it up on exit.
with runloop.devbox.create(name="my-devbox") as devbox:
    devbox.file.write(file_path="main.py", contents="print('hello from a sandbox')")
    result = devbox.cmd.exec("python main.py")
    print(result.stdout())

Limits

  • Deploy to VPC is part of the Enterprise plan, priced by contact.

Strengths

  • Python SDK repository is licensed MIT.
  • Offers SDKs for Python and TypeScript, a CLI and API access.
  • Usage rates are published per CPU-hour, memory GB-hour and storage GB-hour.

Pricing

PricingBasic: free plus usage; Pro: $250 per month plus usage; Devbox CPU $0.108 per CPU-hourPricing page: plans and usage rates, read 16 Nov 2026
TrialPro trial with $50 usage credit, no credit card requiredRead 16 Nov 2026

As written on Runloop's own page on the read date. Prices change: check there before you pay.

  1. Isolated Linux sandboxes for running agent code through Python or JavaScript SDKs.

    Code execution
  2. Open-source sandboxed code execution system with an HTTP API.

    Code execution
  3. Hardware-isolated microVM sandboxes for agent code, local-first.

    Code execution
  4. Run untrusted or agent-written code in isolated Linux sandboxes on Workers.

    Code execution

Last check: . Method: public repository and docs read automatically, page and day written down. Next check due within 90 days. How we check

Is this your tool? Write to hello@agent-tools.directory from an address on its domain to correct a pin or a fact.