Orchestrator Discovery

Surface current work state from MCP Task Orchestrator before taking action.

Workflow

•Get current repository/directory name
•Query projects matching that name
•If not found: prompt user to create project, stop here
•If found: report status, get next task
•Fetch the next task with graph context (single call resolves both)
•Present project status, next task, and graph context to user

MCP Tools

Search for project by repo name:

code

query_container
  operation: "search"
  containerType: "project"
  query: "<repo-name>"

Get project overview with features/tasks:

code

query_container
  operation: "overview"
  containerType: "project"
  id: <project-uuid>

Get next available task:

code

get_next_task
  projectId: <project-uuid>

Fetch task with sections and graph context (single call):

code

query_container
  operation: "get"
  containerType: "task"
  id: <task-uuid>
  includeSections: true
  includeGraphContext: true

The includeGraphContext parameter automatically reads the task's Context Files section, extracts file paths, matches them against atom glob patterns, and returns the hierarchical graph context in the response. No separate query_graph call is needed.

Graph Context Rules

•Only request graph context if a next task exists.
•If the task has no Context Files section or no matching atoms, graphContext will be absent from the response — this is normal.
•If graphContext contains unmatchedPaths, mention them — these are files the task will touch that no atom covers yet.
•If graphContext contains orphanAtoms, note which atoms lack a molecule grouping.
•Do NOT write to the graph during discovery. Discovery is read-only.

Output

If project found with graph context:

code

Project: <name> (<status>)
- Feature A (ACTIVE): 3/5 tasks done
- Feature B (NEW): blocked

Next task: <title>

Graph context:
  <Molecule>: <molecule knowledge summary>
    <Atom>: <atom knowledge summary>
      matched: file-a.ts, file-b.ts
      related: <other atom> — <reason>
  Orphan atoms: <atoms with no molecule>
  Unmatched files: <files no atom covers>

Shall I start on this?

If project found without graph context (no Context Files or no graph data):

code

Project: <name> (<status>)
- Feature A (ACTIVE): 3/5 tasks done

Next task: <title>
Shall I start on this?

If no project found:

code

No project found for "<repo-name>".
Create one to start tracking work?