This is NOT chat. A team of agents, ready to work.

Chat is the tip of the iceberg

One Agent works a task. A team of Agents works the whole job.

✦ Agents pass work to each other — research → analysis → packaging → delivery.

Sapience uses an Agent-as-Tool pattern: any Agent can call other Agents as sub-tools. Ask the Sapience Agent a question, and behind the scenes the Web Research Agent goes out to the open web, the Data Analyst picks up the result, the Coding Agent spins up a sandbox to build the chart, and the answer arrives in your inbox. Consumer chatbots stop at "here's a paragraph." Sapience hands you the finished work.

Agents work on Projects, Goals, Tasks and Meetings — the same objects you do. So “close out the Q3 launch” lands as one instruction, not as a 14-step prompt translation. The depth that makes this possible lives on the Design Philosophy page — the outcome is that an Agent gets work done in your terms, on your priorities, at the speed of a colleague.

Email an Agent. Get work back.

Every Agent has an inbox. You already know how to use it.

✦ web@, data@, [email protected] — and one for every custom Agent you build.

Most Agents in Sapience are also Email Agents. Forward a deal memo, an attachment, a question — the right Agent picks it up, does the work, and emails the answer back. Custom Agents get auto-assigned addresses; Org Admins can choose readable aliases like [email protected]. The interface is the email you already use.

When your colleague is in New York, you don't change how you work — you email them. Sapience Agents work the same way. Email is the universal scaffolding around every executive's day; Sapience Agents plug into that scaffolding instead of asking you to learn a new app.

Not 1:1 chat. Group chat with humans + agents.

Your team thread. Now with experts in it.

✦ N humans + M agents in the same Conversation — full history, per-claim citations.

Sapience Conversations are multiplayer by design: any number of humans and any number of Agents collaborate on the same thread, with full history and persistent context. Add a Research Agent to your customer-success thread. Loop the Translation Agent into a deal review. Save the whole conversation as a Note when it’s done, export to Word/PDF, or share via password-protected public link. "We make working with Agents just like working with humans in a virtual/remote setting" (founder voice).

A workforce already hired

35+ expert Agents and skills, on day one.

✦ Web Research, Data Analyst, Coding, Support — plus org-scoped and user-scoped Agents.

You don't start from an empty chat. Sapience ships with a curated catalogue: the Web Research Agent (Google + OpenAI + Perplexity + Claude search, routed per query), the Data Analyst Agent (a data scientist 24×7, calling the Coding Agent in a sandbox), the Sapience Agent (your concierge — recommends which Agent to use), the Coding Agent (sandbox code execution, visual generation, small web apps) and more. Org Admins can publish org-scoped Agents to everyone; users build their own. Pro-tip from the FAQ: "Projects are Agents." Create one and you've built an Agent that's an expert on YOUR documents — without writing a line of code.

Every frontier model — picked per job

OpenAI · Anthropic · Google · Perplexity. Native SDKs. Routed automatically.

✦ ChatGPT is OpenAI. Claude is Anthropic. Gemini is Google. Sapience is all of them.

Every Sapience Agent runs on whichever lab’s frontier model is strongest for the task — and uses each one via its native SDK (not a thin wrapper). Workhorses (GPT-5, GPT-4.1, Sonnet, Gemini Pro, Sonar) for fast turns. Reasoners (o3, o3-Pro, Opus) for slow-think jobs. Vision and audio routed to whichever lab leads. You get the best of every lab without running an AI team — and zero single-vendor lock-in.