# What feeds a workspace The data sources and integrations behind a client workspace: URL crawl, written brief, document uploads, live web and news search, a read-only profile API and an MCP server. Published: 2026-09-17 ## Public data, several doors Everything a workspace knows about a client comes from public sources. That is a deliberate rule, not a limitation: it keeps the analysis free of internal bias, safe to show the client, and honest about what a buyer or an AI assistant can actually see. Within that rule there are several ways to feed a workspace, and each fills a different gap. ## The URL crawl The usual starting point. Give the workspace the client's domain and it reads the public site: pages, products, pricing, case studies, the about page, the blog. Before it starts, it runs a crawlability pre-check. Sites that block crawlers, render everything in the browser, or hide their content behind scripts are flagged up front, so you know whether the baseline will be thin before you wait for it. The same check is useful later, because a site that is hard for Haycion to read is hard for an AI assistant to read too. Competitors are crawled the same way when they are found or added, so the [Strength Matrix](/what-it-does/competitive-benchmarking-strength-matrix) compares like with like. ## The written brief Some clients have no site worth reading yet. Early-stage companies, new business units, and firms whose site is a single page with a contact form all fall into that category. For those, a written brief starts the workspace instead: what the company does, who it serves, what it claims. The brief goes through the same structuring as a crawl, so the baseline that comes out has the same shape. A brief is also how a requirements-led [vendor comparison](/what-it-does/market-opportunity-sizing-and-prioritisation) starts: describe what a buyer needs, and the workspace finds and scores the vendors that meet it. ## Document uploads A pitch deck, a product one-pager, a market report the client commissioned, a page saved from a partner's site. Uploads in PDF, Word, PowerPoint and HTML can be added to any workspace and are read into its knowledge. They stay in the workspace they were added to. Nothing uploaded for one client is visible from another. ## Live web and news search A baseline is a snapshot. The workspace keeps it current with live web and news search: what competitors announced, what changed in pricing, who raised money, which persona a market report just described. That feed is what writes the daily brief and keeps the Opportunity Graph moving as the field moves. Ask the workspace's agent a question and it can search as well as recall. ## The read-only profile API For consultants who want a workspace's profile inside their own tools, there is a read-only profile API. You create an API key in the workspace, and revoke it there. The API exposes the profile, not the whole workspace, and it cannot write. It is the same shape as the profile the client's agent publishes, so what your tools see is what an AI assistant sees. ## The MCP server Haycion runs an MCP server, so Claude, ChatGPT and any MCP-compatible client can use the workspace directly from a chat: research a company or a competitor, ask the advisor about a finding, start a vendor comparison. Sign-in is by standard OAuth from the client, and each tool has its own scope, so an assistant gets only what it was allowed. The [MCP connect page](/mcp) lists the tools and how to add the server. ## What is not connected yet CRM connections are in early access, listed on the [pricing page](/pricing) under Enterprise. Until they ship, a workspace does not read from or write to a CRM. Saying so here matters more than implying otherwise: a buyer evaluating integrations should know exactly what is live. ## What the client receives - A baseline built from their own public footprint and the open web, with the sources visible. - A workspace that keeps itself current from live web and news search. - Their uploaded material folded into the analysis, and kept to their workspace. - A profile their own tools, and AI assistants, can read through the API and the MCP server. ## How it shows its work Every source is on record. A crawl lists the pages it read and the ones it could not. A brief is kept as the input it was. An upload is listed by name in the workspace it belongs to. A search result that changed a finding is cited on that finding. When you show the client a conclusion, you can show them where it came from, and it is always something they could have read themselves. ## What a result looks like For a client with a readable site, the crawl alone produces a full baseline. Add the deck they sent you and the workspace picks up the two personas the site never mentions. Turn on the daily brief and the first competitor pricing change lands in your inbox the next morning. Connect an assistant over MCP and you can ask "what did we learn about their biggest competitor this week?" from the chat you already have open. ## Questions consultants ask ### Does Haycion connect to my CRM? Not yet. CRM connections are in early access as part of the Enterprise tier. Today a workspace is fed by a URL, a brief, uploads, live web and news search, the profile API and the MCP server. ### What file types can I upload? PDF, Word, PowerPoint and HTML. Uploads stay in the workspace they were added to. ### What is the profile API for? Reading a workspace's profile from your own tools, with an API key you create and can revoke. It is read-only. ### What can an AI assistant do over MCP? Research companies and competitors, ask the workspace's advisor questions, and run vendor comparisons, from a chat. See the MCP connect page for the tool list. ### Does research on one client feed another? Only through the public web. What one workspace learns about a competitor from public sources is available when another workspace looks at the same competitor. Nothing private to a workspace crosses over. --- Canonical: https://haycion.com/what-it-does/what-feeds-a-workspace Source: Haycion, https://haycion.com/