Managed agents, as API

Prompt toproduction-readyagents in seconds

Describe the agent you want to build. Polpo turns it into a portable agent definition with tools, memory, tasks, and sandboxed runtime behind an API.

CodexClaude CodeCursorWindsurf

Trusted by teams shipping agents

Lumea
Userlens
Insurtech Solutions
Open MagiOpen Magi
SuperlogSuperlog
Wordmark
Lumea
Userlens
Insurtech Solutions
Open MagiOpen Magi
SuperlogSuperlog
Wordmark
Lumea
Userlens
Insurtech Solutions
Open MagiOpen Magi
SuperlogSuperlog
Wordmark

Everything behind the API

Use the same agent API for chat, API calls, background tasks, schedules, and long-running workflows.

logs

support-agent

00:01chatstream reply
00:04apireturn result
00:09taskqueue work
09:00schedulerun agent
same agentviachat · api · tasks · schedules

Change behavior without breaking your app

No black box. Prompts, tools, policies, memory, and loops stay reviewable and versioned, so teams can test, approve, and roll back behavior changes without touching product code.

Built-in sandboxing

Polpo spins up a sandbox automatically when an agent needs tools, files, browser, or code access. Simple calls stay lightweight, so you avoid paying for idle compute upfront.

browseacme.com/pricing
writepricing-scraper.ts
runnode pricing-scraper.ts
writepricing.json

Use the same agent across every modality

Stop forcing one model to do everything. Your agent can use the best provider for text, vision, audio, video, or media generation, while your product still calls one stable API.

visionaudiotextvideoTEXT

Keep agent costs predictable

Track spend by agent, model, user, and run. Set limits, compare usage, and keep large agent fleets under control.

monthly spend

$431

budget

$800

54% used$369 left

by model

Claude Sonnet 5$184
Grok 4.20$147
Qwen3 Coder$100

by agent

support-agent$248
research-agent$132
qa-agent$51

Hosted agents at scale

Handle spikes, scale to zero

Elastic scaling for AI-native infrastructure.Polpo runs work when it appears, absorbs traffic bursts, and idles the runtime when nothing is happening.

API & CLI first

Drop-in integration.
Any stack.

Keep your stack and workflow. Your app, CLI, or coding agent calls the same cloud API.

APIs, CLIs, MCP
POST
/v1/chat/completions
works with any framework
Next.js
React
Vue
Svelte
Nuxt
Angular
React Native
Flutter
TypeScript
Python
Node.js
Go
Rust
Swift
Kotlin
Polpo
agent runtime
01
chat
agent session
streaming now
02
12
scheduled task
next in 4m
03
8
queued work
3 retries
running agents3 live
social-monitorsandboxtask
reply-draftersandboxtask
escalatorchat
I found the order.
Checking payment status...
The refund was already issued.
scheduled tasksnext 4m
billing-sync
every 15m
inbox-triage
09:00 daily

No vendor lock-in

Run on any model

Every agent runs on the provider you choose. Bring your own key or use managed credits, and switch without rewriting a thing.

OpenAI
Anthropic
Gemini
Grok
Mistral
Meta
DeepSeek
Groq
Cerebras
OpenRouter
Minimax
HuggingFace
Bedrock
VertexAI
Azure
Copilot
Kimi
Z.ai
Any provider, behind one APIBYOK or managed creditsSwitch models without effort

· faqs

Questions, answered.

The short version. For deeper answers, the docs are exhaustive.

read the docs

Polpo is the cloud backend for agentic products. Define the agent, deploy the harness, and call it from your app, CLI, or coding agent.

$npx @polpo-ai/cli create

Your first agent,
live in 60 seconds

Launch once. Scale from one agent to thousands.

Free tier · No credit card