Agent actor ops

Every accept this package exposes, sourced from the static accepts declarations.

AgentsRoot (system.agents)

From AgentsRoot.ts:93-107:

Op	Description	Required args
$prompt	Create or continue an agent session for a caller mount	(caller-supplied prompt + correlation IDs)
$eval	Evaluate omega-lisp in the persistent agents REPL session	source code
$children	List child actors in this package	—
$status	Package health + storage stats	—
$sessionList	List sessions for a target key	targetKey?: string
$sessionDelete	Delete a session by ID	sessionId: string
$sessionClearAll	Delete all persisted sessions and active child agents	—
$sessionSeed	Seed a session transcript for tests/tools	(transcript)
$metrics	Return system.agents metrics	—
$traceQuery	Return activity frames for a session/request	sessionId: string, requestId?: string
session.create	Create or ensure an agent session record	sessionId: string, targetKey: string, principalId?: string
session_state.upsert	Persist JSON session state by key	sessionId: string, key: string, data: unknown
session_state.read	Load session history for a session ID	sessionId: string

AgentsRoot.isPromptRouter = true exempts the root from the standard $prompt seal.

MemoryActor (system.agents.memory)

From MemoryActor.ts:86-102:

Op	Description
memory.save	Save a memory record for an actor (mount, content, tags?, targetKey?, profileKey?, scope?)
memory.get	Get a specific memory by ID (id: number)
memory.search	Search memories — hybrid: vector + FTS5, degrades gracefully (mount, query, limit?, minScore?, askSources?, targetRef?, targetKey?, profileKey?, sessionId?)
memory.history	Recent memory records for a mount (mount, limit?)
memory.searchTier	Which search tier is active (hybrid / fts5 / like)
memory.consult	Ask the actor to answer from its own memory context (question, context?, maxTokens?)
plan.create	Create a plan (mount, name, steps)
plan.update	Update a plan step (planId, stepIndex, status)
plan.active	Active plan for a mount (mount)
plan.complete	Mark a plan complete (planId)
plan.history	Plan history for a mount (mount, limit?)
memory.audit-trail	Memory access log for a mount (mount, limit?)
plan.execute	Start step execution (planId, stepIndex, ownerMount)
plan.execution-complete	Mark step execution result (executionId, status, output?)
plan.executions	Active executions for a plan (planId)

EmbeddingService (system.agents.embeddings)

From matrix.json and the service source:

Op	Description
embed.generate	Generate an embedding for given text
embed.batch	Generate embeddings for an array of texts
embed.search	Vector search — given a query embedding, return ranked memory IDs
embed.providers	List configured embedding providers (OpenAI, Gemini, local)
embed.status	Active provider + cache statistics

Manifest summary

// projects/matrix-3/packages/agents/matrix.json
"exports": ["memory", "embeddings"],
"storage": { "sqlite": { "domain": "agents", "schema": "schema.sql", "version": 2 } },
"permissions": { "fsPolicy": "project-readwrite", "network": true, "subprocess": false, "env": false }

See also

• Session schema
• Tool schema
• Memory section

Source: All op rows are read verbatim from static accepts in the corresponding actor source files. The package manifest is projects/matrix-3/packages/agents/matrix.json.