Help centre
How can we help?
74 guides and reference pages, taken from the product itself. Search, or start with a topic.
Start here
From sign-up to an agent answering on your website, in order.
- Step 1Quickstart: your first agent in ten minutesSign up, create an agent on your own model key, add knowledge, test it in the Playground and stream an answer from the API.3 min read
- Step 2Add your LLM key (bring your own key)Agents only answer once they have your own OpenAI, Anthropic, Google Gemini or Groq key. Where to get one, where to save it, and how it's protected.3 min read
- Step 3Create an agentMake a new agent from a template or from scratch: name, provider, model, your provider key and a system prompt.2 min read
- Step 4Install the widget on your websiteCopy the embed code from the agent's Deploy tab, add your website to Allowed websites, and paste the tag before </body>.2 min read
Browse by topic
Everything about one part of SentientOne, in the order you'll need it.
What SentientOne is, the words it uses, and your first agent answering through the API in about ten minutes.
All 4 articlesCreate an agent, set its model and behaviour, write its prompt, test it in the Playground and keep track of every change.
All 6 articlesThe providers and models agents can use, how your own provider keys are stored and used, and how to pick a model.
All 3 articlesGive an agent your documents, websites, snippets and Q&A, understand how hybrid search and rerank find the right passage, and load knowledge by API.
All 7 articlesChain agents, classification, extraction, guardrails, MCP tools, HTTP, Slack and email on a canvas, then run them by webhook or API.
All 8 articlesPut an agent on any website with one script tag, a publishable pk-so- key and a list of allowed websites.
All 6 articlesThe REST API: keys, authentication, the streaming chat endpoint, every event and error code, rate limits, request ids and the workflows API.
All 13 articlesConnect agents to your Model Context Protocol servers so they can call your APIs in the middle of an answer.
All 5 articlesEvery request traced end to end, with retrieval, model rounds, tool calls and tokens, plus analytics per agent.
- Use Tracing to see every request
- Trace reference: spans and attributes
- Analytics: requests, errors, latency and tokens
How credits work, what each action costs, buying packs, auto top-up, what happens at zero, and receipts.
All 5 articlesSigning in, two-step verification, your settings, deleting your account, what data is kept, and how to reach support.
All 6 articlesFixes for the widget, knowledge that isn't used, MCP servers that won't connect, API errors and website crawls, plus the FAQ.
All 7 articles
Popular
- POST /v1/chat/streamAPI reference
- Install the widget on your websiteWebsite widget
- How retrieval works: hybrid search, fusion and rerankKnowledge and RAG
- Errors and error codesAPI reference
- Connect an MCP server to an agentMCP tools
- How credits workCredits and billing
- Run a workflow by webhook or APIWorkflows
- The widget doesn't show or doesn't answerTroubleshooting and FAQ
API reference
- API overviewWhat the SentientOne API does, its base URL, the endpoints you'll use and how requests, streams and errors fit together.
- API keys: secret and publishableThe two kinds of SentientOne key, where to find them, where each one may be used, and what to do if a secret key leaks.
- Test requests on the API keys pageSend a real request to any of your agents from the app, watch the events stream in, and copy ready-made cURL, TypeScript or Python.
- AuthenticationThe headers every API call needs: X-Api-Key and X-Agent-Id, plus the optional X-Llm-Key, X-Request-Id and X-Perf-Timing.
- POST /v1/chat/streamThe chat endpoint field by field: request body, limits, what the agent does with each request and how the streamed reply ends.
- Streaming eventsEvery server-sent event the chat stream can send, in order, with each field, plus how to parse the stream safely.
- Conversations and historyHow conversation_id works, how much history the model sees, and where conversations from every channel show up.
- Errors and error codesThe error envelope, every error code the chat, widget, workflow and ingest endpoints return, its HTTP status and what to do about it.
- Rate limitsPer-agent requests per minute, widget limits per key and per visitor, workflow limits, the X-RateLimit headers and how to retry.
- Request ids and correlationSend X-Request-Id and follow one request through the API, its trace, error reports, your model provider, MCP servers and webhooks.
- Latency and performanceMeasured timings for a reply, what SentientOne does to keep them low (caches, hedged calls, retries) and how to time your own requests.
- Workflows APIStart live runs of published workflows with POST /v1/workflows/run or a webhook, read step events, and handle run errors.
- Code examplesCopy-ready examples: cURL, a Node stream reader, a Next.js route that keeps your key server-side, and async Python with conversations.
Still stuck?
Email us with what you tried and any request id. A person replies.