SentientOne Documentation

A simple, request-driven agent platform. You build an agent in the dashboard — give it a system prompt, a persona, the tools it can call, and the knowledge it should ground its answers in — then invoke it from any internal app with a single POST /v1/chat. One endpoint, two headers, JSON in / JSON out. That's the whole product surface.

• Prompt + personaConfigure the system prompt and tone once, version it, and roll back when you need to.
• Tool callingConnect MCP servers so the agent can look things up in your APIs (orders, products, tickets, anything you expose).
• KnowledgeUpload documents, drop in Q&A snippets, crawl a website — the agent retrieves the right context on every call.
• One API callSame call shape from every app — web, mobile, internal tools, cron jobs. Route to a different agent by changing the X-Agent-Id header.

We're deliberately not in some of the spaces other "agent platforms" live in. Knowing what we don't do helps you pick the right tool for your problem.

• Not an always-on agentAgents don't sit in the cloud running on their own. They wake up when your app calls them, do their work in a few seconds, and shut down. If you need long-running autonomous workers, that's a different shape of product.
• No idle billingWe don't charge for time agents sit idle or for requests you don't send. The plan covers a monthly request bucket; if you don't use it, you don't pay for it.
• You pay the LLM directlyToken cost goes straight to your provider (OpenAI, Anthropic, Gemini, Groq) on your own key. We never mark up tokens or proxy your provider bill — see LLM Keys.
• Not a workflow orchestratorOne request, one conversation, one reply. We're not a long-running pipeline engine — if you need DAGs of steps that run for hours, pair us with a workflow tool and call us per step.
• Not a model trainerYour prompts and conversations are never used to train any model. The platform passes data through to the LLM and back — see Security.