01
Before: procedures exist, but finding the current answer is slow
- People repeat the same lookup because the approved source and policy owner are hard to find.
- A sensitive exception still needs a person; a chat answer is not an approval.
Illustrative example
A governed policy chat for repeated internal questions
This illustrative workflow shows a standalone Prostir Agent built from approved procedure files. Knowledge search is enabled only after the resources are ready; access uses OAuth grants for verified exact email addresses or a normalized exact company domain. A supported AI client may use the Agent only when the user's account, connector, transport, and authentication support it.
Before you read
What gets published
After successful validation and publication, the Agent's actual saved slug follows the patterns https://{slug}.ai.prostir.build and https://{slug}.ai.prostir.build/mcp; these are non-live patterns, not this example's URL. Access is limited to explicit OAuth grants. Source and version metadata can accompany answers only when that knowledge build and citation data are ready and enabled; no exact-quote, freshness, adoption, or Slack-reduction result is promised.
Best for
Operations, people, IT, and security owners who need controlled answers from internal procedures while keeping review and exceptions with people.
Where the work happens
Illustrative standalone Agent · Approved resources · Explicit OAuth grants · Rebuild and republish updates
02
Illustrative workflow in Prostir Studio
This is a product walkthrough, not a customer implementation or measured result.
- 01Select and approve the procedure files to use.
Upload approved exports or files as Agent resources; no direct external-drive connector is assumed.
- 02Build knowledge, review it, then enable search.
Source and version metadata is shown only when the build and citation data are ready and enabled.
- 03Create exact OAuth grants.
Allow verified exact email addresses or one normalized exact company domain—never suffix matching or an unlisted link.
- 04Publish the standalone Agent.
Its canonical address follows https://{slug}.ai.prostir.build and https://{slug}.ai.prostir.build/mcp; an external AI client also needs a supported account, connector, transport, and authentication.
03
Approved knowledge sources
Use only resources an owner approved for this Agent.
- docapproved-procedures/Versioned procedure exports with a named owner and review date.
04
Illustrative conversation in a supported AI client
Illustrative response, not a live customer transcript or exact-quote guarantee.
Illustrative response, not a live customer transcript or exact-quote guarantee.
Where is the current on-call procedure, and who approves an exception?
Illustrative answer: search the approved knowledge, show source/version metadata only if enabled, and route any exception to the named human owner.
05
What this setup can change
- It can reduce repeated lookups, but it does not guarantee fewer Slack messages or company adoption.
- After a source changes, rebuild the knowledge and republish before treating the update as available.
- This page describes a standalone Agent. Shared tasks, files, and collaboration belong to the separate Company product on the applicable custom plan.
Solutions
Internal AI knowledge base with the right ownership model
Choose a standalone Agent when one owner controls the assistant, or scope a private Company foundation when members need to update Tasks, Data, Files, Knowledge, and allowed attachments together.
Solutions
Choose the ownership model before the tool
Tell us who answers, who edits, and what must remain private. We will help you decide between a standalone knowledge Agent and the Company foundation.