Sapiom overview
Sapiom is a platform for building, deploying, and operating observable agents. Create an agent with Agent Studio or Sapiom MCP, then run it with managed capabilities, durable execution, and one trace of what happened.
Sapiom Router is the LLM-routing product. Call it directly from an application, or use it during an agent run when the model request should be part of the agent’s execution history.
What do you want to do?
Section titled “What do you want to do?”The two product paths
Section titled “The two product paths”Build and operate agents
Section titled “Build and operate agents”An agent is a controlled, multi-step automation. Its deployed steps can use Sapiom capabilities such as web search, browser automation, compute, data, files, images, and audio.
Use Agent Studio when you want an integrated workspace with a coding session, Canvas, local testing, deployment, and run inspection. Use Sapiom MCP when you want Claude Code or Codex to perform the same project and cloud-lifecycle operations from its own terminal or editor experience.
Both interfaces produce the same agent project and deploy to the same runtime. They are two ways to author and operate an agent, not two different capability systems.
Route model traffic
Section titled “Route model traffic”Use Sapiom Router when you need LLM inference. It accepts provider-compatible requests, selects the available model infrastructure behind a model label, and records the request as Router traffic.
Router can be called directly with a Sapiom account credential. When an agent calls Router during a production run, Sapiom can additionally associate the model activity with that run and step.
Start with the mental model
Section titled “Start with the mental model”Authentication at a glance
Section titled “Authentication at a glance”You need a Sapiom account before using Agent Studio or Sapiom MCP, including for local project creation, editing, checks, and Local Run. Authentication identifies who is using the authoring tools; it does not turn local work into a cloud agent run.
After you sign in, local checks and Local Run still execute on your machine. Local Run replaces Sapiom capability calls with stubs, so it makes no real Sapiom capability request and creates no Sapiom capability spend. Linking, deployment, production runs, and triggers additionally operate organization-owned cloud state.
Your coding-agent account is separate from your Sapiom account. Claude Code uses your Anthropic authentication; Codex uses your OpenAI authentication. You need both the coding-agent authentication and Sapiom authentication for an Agent Studio or Sapiom MCP authoring session.
Where to go next
Section titled “Where to go next”© 2026 Sapiom, Inc.