agenticonsult logoagent i /consultDocs
Features

Context Artifacts

Author curated context packages, attach a knowledge graph at a chosen depth, and ground agent sessions and terminal workspaces with them.

The Context Artifacts view (sidebar ▸ Data) is where you build and manage context packages — the curated notes, file references, and optional attached graphs that ground agent sessions from their first prompt. This page is the hands-on guide; why artifacts matter and how grounding behaves is Context artifacts & grounding.

Author an artifact

Create an artifact and give it a clear name — sessions will be grounded with it by that name. The editor works in three modes, switchable at the top:

  • Visual — the structured builder: a description, freeform notes, and the artifact's file references. Add references to individual files or whole folders; you can also drag a file straight from your file manager onto the artifact, and it is added as a member.
  • Raw — the artifact as plain Markdown, edited directly. Everything the visual builder produces is ordinary text; nothing is hidden behind the UI.
  • KG — the artifact's optional attached knowledge graph (next section).

Edits save automatically as you type — there is no save button to remember.

What belongs in a good artifact: the goal, the decisions already made, the constraints, the vocabulary, and pointers to exactly the files that matter. Write it as the briefing you would otherwise repeat.

Attach a knowledge graph

Switch to KG mode to bundle a purpose-built graph with the artifact, built from its file references. Pick the build depth — the same three choices as everywhere in the platform:

ChoiceWhat the agent getsCost
Structure onlyThe folders and documents as a browsable mapnone — no AI involved
Structure + EntitiesPlus extracted concepts and relationships to traverseyour own agents' reasoning, no external API
Full (Entities + Vectors)Plus semantic search over the materialyour Gemini key, checked before the build starts

Structure only is a complete, useful state — start there for most artifacts. The choice is recorded on the artifact, and it tracks reality: enriching the graph later moves the badge to entities on its own. The attached graph appears in Cortex like any other workspace, and the same documents stay single nodes in the one knowledge graph — attaching them here duplicates nothing.

Preview what the agent receives

Before you trust an artifact with real work, open its preview: the exact materialized fragment a grounded session would be handed — notes, resolved file references, and graph pointer included. If the preview reads like a good briefing, the grounding will too. This is also the fastest way to catch a stale reference or a note that no longer applies.

Ground sessions with it

Artifacts do their work when attached where sessions launch, in the Terminals view:

  • Attach to a terminal workspace to make the artifact the default context for every agent pane in it — the right move for a fleet: attach the effort's briefing once, and every worker inherits it.
  • Attach to a single pane for one specialist that needs its own briefing, layered on top of the workspace default. A session's effective context is the combination of both, deduplicated.

Back in the Context Artifacts view, each artifact shows where it is in use — an "attached in" marker per terminal workspace, with a jump straight to that workspace.

Two properties worth internalizing:

  • Grounding is captured at launch. Editing an artifact does not reach into sessions already running — respawn the pane (or start a fresh session) to carry updated context. A live agent's briefing never shifts mid-task.
  • Context only, never capability. A grounded session keeps its full toolkit, model settings, and permissions; the artifact changes what it knows at start, not what it can do.

Every door uses the same machinery: sessions you launch, sessions Voice stands up, and sessions agents spawn for other agents all resolve artifacts through one backend — an agent-dispatched worker inherits the workspace's context exactly as an operator-launched one does.

Artifacts are files

Every artifact lives as a Markdown file under documents/context-artifacts/ in your workspace. That has two practical consequences: your agents can read and even maintain artifacts as part of their work, and any edit made outside the view — by you or by an agent — shows up in the view automatically. Your version-control and backup habits cover them like any other document.

Where to go next

On this page