$4 per connector · up to 4 apps · Up to $16
Bring your own key or use our model router
Connect your Claude or Codex subscription. We charge $0 for those tokens. Or use any model through our model router, at 0 markup.
Connect your planChange provider mid-session
Hit your Claude Code limit, or want a different model on the job? Hand the task to Codex, Hermes or another agent. It picks up where the last one left off.
Start a taskAdd-ons. One click to install.
Give your agent a tool, an app or a chat channel. It works the same in the browser and through the API.
Tools
Search the web, browse a site, make images, video and speech. One click and any agent can use it.
Browse toolsConnectors
GitHub, Gmail, Notion and 2,000 more apps. Sign in once; every agent can use them.
Browse connectorsChannels
Slack, Discord, Telegram, WhatsApp, iMessage. Your agent answers in all of them.
Browse channelsUse any agent via API.
Use any agent via one API. Use it, or build your own product on top.
Build multi-agent flows
Let one agent plan, another write the code and another test it, all in one flow that runs on its own.
Agent as backendBuild your own agent product
Pick a harness, add your prompt and tools, and ship it under your own name. Point your coding agent at our docs and it wires this up for you.
Read the API docsBuilt on the Agent API
450,000+ agents launched on the API. The Agent API powers these products. Every agent their users talk to runs on AgentSky.
All case studiesBuilding a software factory where the AI CEO routes tasks to Claude Code, Codex, and Hermes through one API.
Prewarms a cloud agent so the first user prompt lands on a ready worker. No cold-start delay in the build path.
Moved from Claude Code to pi and open-source models with one config change. The product API and backend stayed put.
Bring your own key or use our model router
Connect your Claude or Codex subscription. We charge $0 for those tokens. Or use any model through our model router, at 0 markup.
Connect your planChange provider mid-session
Hit your Claude Code limit, or want a different model on the job? Hand the task to Codex, Hermes or another agent. It picks up where the last one left off.
Start a taskAdd-ons. One click to install.
Give your agent a tool, an app or a chat channel. It works the same in the browser and through the API.
Use any agent via API.
Use any agent via one API. Use it, or build your own product on top.
Build multi-agent flows
Let one agent plan, another write the code and another test it, all in one flow that runs on its own.
Agent as backendBuild your own agent product
Pick a harness, add your prompt and tools, and ship it under your own name. Point your coding agent at our docs and it wires this up for you.
Read the API docs