architect
Ask when consequential design choices must be settled before or during implementation — investigates context, compares…
Sprout can run child agents in separate processes coordinated by a WebSocket bus. This keeps long-running or async subagents isolated while preserving one session-wide event stream.
How it fires
How this agent gets triggered: by you, by Claude, or both.
Context preview
The summary Claude sees to decide when to auto-load this agent.
Sprout can run child agents in separate processes coordinated by a WebSocket bus. This keeps long-running or async subagents isolated while preserving one session-wide event stream.
Sprout can run child agents in separate processes coordinated by a WebSocket bus. This keeps long-running or async subagents isolated while preserving one session-wide event stream.
Bus server: local WebSocket pub/sub server.
Bus client: connects runtime processes to the server.
Spawner: starts or coordinates agent processes, tracks handles, and supports blocking, non-blocking, and shared-agent workflows.
Agent process: receives a start message, constructs an agent, runs it, and sends events/results back over the bus.
Source of truth:
The bus is bidirectional. The host sends commands down to child processes and receives events/results back. WebSocket fits this better than one-way SSE.
The server binds locally on a runtime-selected port. The CLI starts bus infrastructure before constructing the session runtime.
Source of truth:
The session controller subscribes once to session events and relays child process events to the UI bus. This avoids per-child subscriptions and lets the TUI/web client show a single event stream.
Source of truth:
Non-blocking delegates return handles. The caller can wait for a handle or send messages to a shared handle. Resume can pre-register completed handles from previous logs so an agent can continue without rerunning completed work.
Source of truth:
Child agents receive the root surfaced memory block so repeated subagents do not each redo recall. Archivist is intentionally excluded so it investigates memory from a clean context.
Source of truth:
A self-improving, multi-agent AI system that decomposes tasks and delegates to specialist agents. Sprout is an autonomous coding assistant built on a recursive agent architecture — where every action is a goal-directed delegation to a subagent.
Ask when consequential design choices must be settled before or during implementation — investigates context, compares…
Memory investigation, synthesis, and authorized curation when surfaced memories are insufficient
Systematically diagnose and fix bugs: find root cause before attempting any fix
Observe Sprout's live session behavior and send concise guidance when it is drifting, stuck, or missing an important instruction
Explore and understand unfamiliar codebases, map architecture, and synthesize structural insights
Ask when you need to know what Sprout agents, tools, or MCP servers are available, need a reusable capability plan, or need a new…