Create the first knowledge-base space

A new org has zero KB spaces. Before the team can write articles or an agent can call kb_search, at least one space must exist.

TL;DR

1. kb_list_spaces — see whether a space already exists.
2. If empty: kb_create_space({ name, slug, description? }).
3. (Optional) kb_create_document({ spaceId, title, body, audiences }) to seed a welcome doc.
4. Hand off to skill://kb/import-articles-in-bulk or skill://kb/import-from-google-docs for content.

Step 1 — check current state

{ "name": "kb_list_spaces", "arguments": {} }

If a space already exists, skip to step 3 or stop — the org is onboarded.

Step 2 — create the first space

The space is the namespace agents and humans search within. Slug must be globally unique within the org (lowercase, digits, hyphens, 1–64 chars).

Pick a name + slug that match the audience boundary:

• product / Product — internal: how the product works, internal runbooks, postmortems.
• help / Help center — public: end-user-facing FAQs and guides.
• engineering / Engineering — internal: code conventions, architecture, on-call docs.

{
  "name": "kb_create_space",
  "arguments": { "name": "Help center", "slug": "help" }
}

If creation rejects the slug, it's almost always uniqueness — try a more specific slug (acme-help rather than help).

Step 3 — optional welcome doc

{
  "name": "kb_create_document",
  "arguments": {
    "spaceId": "<space-id-from-step-2>",
    "title": "How we work",
    "body": "... markdown ...",
    "audiences": ["admin"]
  }
}

The body is markdown; chunking and tagging guidance lives in skill://kb/import-from-google-docs.

Step 4 — onward

For more spaces, call kb_create_space again. For bulk content seeding, follow:

• skill://kb/import-from-google-docs — chunking strategy, embedding behavior, FTS performance.
• skill://kb/import-articles-in-bulk — generalized CSV/JSON pipeline.

What NOT to do

• Don't create a "default" or "general" space. Spaces are search-filterable; agents will conflate everything if there's only one. Pick a real audience boundary.
• Don't skip the slug discussion. Slugs are user-visible (in URLs, in agent search filters) and changing them later is awkward — kb_create_space doesn't have an "update slug" tool.
• Don't import content before the space exists. kb_create_document requires a spaceId; bulk loops fail loudly without it.

Related

• skill://kb/import-from-google-docs — chunking + embeddings deep-dive.
• skill://kb/import-articles-in-bulk — generalized bulk pipeline.