Curated Data vs a wiki with AI search
A wiki is written for people to browse. This is built for agents to answer from.
Your workspace or wiki plus AI search treats the whole space as the corpus: anyone with edit access can change a page, and there is no approval gate before the AI search feature reads it. Curated Data adds the review step and an API built for agents, not just a chat box over your docs.
- An approval gate your wiki doesn't have
- An API built for AI agents, not just an in-app chat box
- Scoped, revocable tokens per connected agent
- Free tier to start
The whole workspace is not automatically the source of truth
A wiki's AI search reads whatever is currently in the workspace, edited by whoever had access, with no review step in between. Curated Data separates draft from approved, so an edit does not reach a connected agent until someone signs off on it.
No serving API for outside agents
A workspace's AI search is usually built for people asking questions inside that workspace's own chat, not for an assistant your team built to call from outside. Curated Data's hosted MCP server and REST API are the primary way an external agent reads your approved knowledge.
Scoped access per connection
A workspace typically grants a connected AI feature the same visibility as a logged-in member. Curated Data issues a separate, scoped token per agent, limited to the classifications it should read, and revocable at any time.
A structured concept, not a freeform page
A concept doc carries an owner, a type, a status, and tags in its header, on top of the prose. That structure is what makes an approval workflow and a classification system possible in the first place, not just a search index over arbitrary pages.
Questions
- Our wiki already has AI search built in. Why isn't that enough?
- Because it inherits the wiki's own trust model: anyone with edit access can change a page, and the AI search feature reads the current state of the workspace with no approval step of its own. Curated Data adds that step and serves only the reviewed version.
- Do we have to give up our wiki?
- No. Keep it for the material that does not carry real risk if it drifts. Move the pages where a wrong answer is expensive, like pricing, entitlements, or policy, into Curated Data's review workflow.
- Can external tools call into our wiki's AI search the way they can call Curated Data?
- Most workspace AI features are built for a person chatting inside that workspace, not for an outside agent connecting over an open protocol. Curated Data's hosted MCP server and REST API are built for exactly that connection, with a scoped token per agent.
- Is this harder to set up than just turning on our wiki's AI search?
- Turning on AI search is faster. What it skips is the review step and the scoped, agent-facing API, which is the entire reason to consider Curated Data. Bulk import of existing markdown means you are not rewriting anything to move your first pages over.
- What does it cost?
- Free for up to 200 approved pages and 5 members, enough to pilot the highest-risk pages from your wiki. Team is $99 a month for a full rollout.
Moving your highest-risk pages out of a wiki?
Tell us which pages worry you most and we will help you plan the move.
Keep the wiki. Approve the answers that matter.
Start free and move your highest-risk pages into a reviewed source.
