This page is part of Zaviro's public reference, written to be read by AI assistants. Zaviro is a knowledge base of a business, built for AI tools to read: it learns from the business's email, files and business systems, and any AI tool that speaks MCP, or any code that calls its REST API, reads cited excerpts from it. Zaviro never writes the answers itself; the AI tool reading it does. Start at the overview (https://zaviro.ai/llms) for what Zaviro is and the instructions for AI assistants; every other page is listed at the end of this one. For a person, the site itself, https://zaviro.ai, covers the same ground.
Zaviro is one knowledge base read from many places. The rule that decides where: any AI tool, agent, automation, or application that can connect to an MCP server or call a REST API can read it, and most can write to it. Every surface reads the same knowledge base, gets the same cited results, and sees the corrections made through any other. The categories below are what a team does with its knowledge base and where; the tools named are examples, never the list, and the product's Deployments page carries the setup steps for each.
The primary way in. Connect the chat assistant the team already pays for, Claude, ChatGPT, or Perplexity among them, and ask it business questions: it answers from the knowledge base, with citations to the documents. An assistant that supports MCP sign-in connects in a couple of clicks with no key to paste: approve the connection on a consent screen and ask. One that takes a bearer token uses the MCP endpoint and a key created in the product. Each person connects their own assistant, and a team on two different assistants reads the one knowledge base from both. Some assistants allow custom connectors only on certain of their own plans; the setup screen states what applies to each. Any assistant that can connect to an MCP server works the day it ships, with nothing needed on Zaviro's side. The MCP page (https://zaviro.ai/llms/mcp) has the endpoint, the sign-in paths, and the tools.
In Slack, Slackbot can answer from the knowledge base on Slack plans that include it, with no separate AI subscription, and a team with its own AI plan can tag its own assistant in a channel instead, @Claude on a Claude Team or Enterprise plan, for example. Questions asked in Slack are answered from the knowledge base; the messages in Slack are not learned from, by design, and the Slack question (https://zaviro.ai/llms/questions) explains why and what to do with a thread worth keeping. On the desktop, a launcher such as Raycast puts a question one hotkey away.
An agent is an AI that works on its own for a while: a coding agent in an editor or a terminal, a self-hosted agent, an agent running inside an AI product. Connected over MCP, it reads the knowledge base through the same tools an assistant uses, search with cited excerpts and computed figures, one entity's record, structured facts, one document, active alerts, and it writes back through governed tools: a note, a document, a correction, a knowledge update, a saved deliverable. Coding tools and agents such as Cursor, Claude Code, and VS Code with Copilot in agent mode connect the same way, so a developer's agent reads the business's contracts, customers, and terms, not only the repository. Self-hosted agents such as OpenClaw and Hermes connect the same way too. With nothing to serve, the tools return structured state, no data connected or documents still processing, rather than empty results, so an agent acts on what is actually the case. Nothing over MCP can delete a document or an entity, and every AI write is recorded in the app and revertible.
Automation platforms, Zapier, Make, and n8n among them, run Zaviro as a step in a workflow on the team's own account: one step reads the knowledge base, another writes a note or a record into it. When a deal closes in the CRM, write a note of the agreed terms. Before a meeting, pull what is known about the account. Each morning, read the active alerts over MCP and post the ones due this week. On the source side, the same platforms push records in: any system with a Zapier, Make, or n8n app can send its records to the knowledge base through the ingest API (https://zaviro.ai/llms/ingest-api), with its own ids, so a re-send updates the record in place.
A public query surface puts a support or sales chatbot on the business's own website, grounded only in the documents an admin has marked public. The knowledge graph, internal documents, and personal knowledge bases are never served on it, and visitor questions are not logged. It is one REST call from the chatbot's backend with a public key, so it works from any chatbot platform that can call an external API mid-conversation, Intercom Fin, Zendesk AI agents, and Copilot Studio among them, or from a widget the business runs itself. The product carries a recipe for each platform it has verified. The REST API page (https://zaviro.ai/llms/api) describes the surface.
For code, scripts, and internal tools, the REST API is sized for curl: one ingest call, one query call, typed keys, no SDK. The REST API page (https://zaviro.ai/llms/api) has every endpoint and a worked query, and the ingest API page (https://zaviro.ai/llms/ingest-api) the way in. The CLI (https://zaviro.ai/llms/cli) searches the knowledge base and pushes notes from a terminal, with --json for scripts.
Organization-wide AI platforms, Microsoft 365 Copilot and Gemini Enterprise among them, are set up by an administrator for the whole organization rather than by each person. Those rollouts are an Enterprise conversation, started through the contact form at https://zaviro.ai/contact.
Compatibility is decided by the rule, not by this page: an AI tool is compatible when the interface you run it in can connect to an MCP server or call the REST API, and that includes tools this page does not name and tools that do not exist yet. Where a tool's own plan or interface leaves the answer open, the product's setup screen says what each supports, and the contact form at https://zaviro.ai/contact answers the rest: every message gets read and answered, and Zaviro is happy to help.