Skip to main content
An indexed source is a named collection of organizational knowledge: a handbook, documentation tree, wiki export, knowledge vault, research archive, or repository.

​
What to index

Documentation

Product guides, runbooks, policies, and reference material.

Knowledge bases

Linked notes, internal wikis, and curated research.

Repositories

Code and the documentation that explains how it works.

Document archives

Collections that need semantic retrieval but have no live query API.
Do not index a service merely because it contains text. If it has a useful query API or MCP server, connect it so results stay current and actions remain possible.

​
Source lifecycle

1

Register the source

An admin creates a permanent source slug and a description written for an agent deciding when to search it.
2

Publish the content

Your publishing workflow sends the complete document set. A publish replaces the previous version of that source.
3

Build the index

Port of Context extracts the content, builds the search index, and marks the source ready to serve.
4

Search through the gateway

Agents search one source or merge results across several sources, then read only the documents they need.

​
Write a useful description

The source description is routing context. Name the material, the questions it answers, and any important boundary.
Good source description

​
Place sources on the map

Attach a source to the team, system, or domain it documents. When an agent retrieves that entity’s context, it discovers the relevant source without needing its name in the prompt.
Each publish is the complete source. Files missing from the new publish are removed from the next index.

Search, then read

Next: Skills

Capture repeatable ways of working in a shared skills library.