ClaudePaw
Now in public beta

Autonomous AI agents
for your business

Deploy agents that think, act, and deliver. Real tools, persistent memory, composable skills. Not chatbots — agents that run ReAct loops and get work done.

Start building freeRead the docs
Agent Playground
User: Find competitor prices for roof washing in Trondheim
web_search("takvask pris Trondheim")
read_url("renholdsportalen.no/...")
web_search("takvask Trondheim anmeldelser")
Agent: I found prices from 5 providers in Trondheim. Here's the breakdown...
3 tool calls · 2 iterations · 1,847 tokens · $0.004

Everything agents need to operate

A complete runtime — not a wrapper around a chat API.

ReAct Agent Loop

Agents reason step-by-step, use tools, observe results, and iterate. Not scripted — genuinely autonomous.

Real Tools

Shell execution, browser automation, file I/O, web search, MCP servers. Agents act in the real world.

Composable Skills

Drop in SKILL.md files to teach agents new capabilities. Browse the registry or write your own.

Persistent Memory

Agents remember across conversations. Auto-appended history plus agent-curated notes.

Subagents

Spawn background agents for parallel work. Shared workspace, independent context windows.

Multi-Tenant

Isolated orgs, API keys, plan-based tool access, usage metering. Built for SaaS from day one.

Custom Domains

Point your own domain to the API. CNAME setup, automatic SSL, org-scoped routing.

API-First

OpenAI-compatible chat endpoint plus full agent API. SSE streaming with tool status events.

How it works

From message to autonomous action in milliseconds.

1

Message arrives

Gateway authenticates, resolves tenant, checks plan limits, routes to agent.

2

Context is built

System prompt, active skills, relevant memory, tool definitions, conversation history — all assembled.

3

Agent reasons & acts

ReAct loop: LLM thinks, calls tools, observes results, repeats until the task is done.

4

Results stream back

SSE stream delivers text, tool status, and final stats. Memory updated for next time.

Simple, honest pricing

Pay for what you use. Upgrade when you need more.

Free

$0/mo

Try it out. One agent, core tools.

  • 1 agent
  • 100k tokens/mo
  • Web search & URL reader
  • Memory
  • Messaging
Start free
Most popular

Pro

$29/mo

Unlock the full toolkit. Shell, browser, files.

  • 5 agents
  • 2M tokens/mo
  • All Free tools
  • Shell execution
  • Browser automation
  • File I/O
  • Container compute included
  • $0.01/1k token overage
Start Pro trial

Agency

$99/mo

Unlimited agents. Subagents. MCP. Priority.

  • Unlimited agents
  • 10M tokens/mo
  • All Pro tools
  • Subagent spawning
  • MCP server connections
  • Custom domains
  • Priority container
  • $0.008/1k token overage
Start Agency trial

Ship your first agent today

Free tier. No credit card. From zero to autonomous agent in minutes.

Get startedDocumentation