OpenClaw Subagent Handoff Planner
Plan multi-agent handoff chains — configure announce messages, task delivery visibility, and session bootstrapping — then export ready-to-use YAML config. Based on OpenClaw PRs #81383, #81319, and #81365 merged May 13, 2026.
Why Use the Subagent Handoff Planner
Built from 3 PRs merged into OpenClaw on May 13, 2026 — covering in-process announce, visible task delivery, and session bootstrapping.
Visual Handoff Builder
Define multi-agent topologies and handoff chains through an interactive form. See the complete flow diagram update in real time as you connect agents.
In-Process Announce Handoff
Configure handoff announcements using the in-process announce system from PR #81383. Choose from templates or write custom messages with variable interpolation.
Visible Task Delivery
Control task delivery visibility per handoff (PR #81319). Make subagent task assignments transparent for debugging or hide them for clean user-facing output.
Session Bootstrap
Enable pre-send session initialization per handoff (PR #81365). Bootstrap agent sessions before the first message to eliminate cold-start latency.
Real-Time Validation
Catches config errors as you type — duplicate agent names, self-referencing handoffs, missing conditions, and orphaned agents with no handoff connections.
Live YAML Preview
See your complete handoff config update in real time as you adjust settings. Copy the generated YAML directly into your openclaw.yaml config.
Build Your Handoff Config
Add agents, define handoff chains with conditions and announce messages, then copy the generated YAML into your config.
Agents2
Define the agents in your multi-agent setup. Each agent needs a unique name, a role, and a model.
Agent 1
Agent 2
Handoff Chains
Define how agents hand off tasks to each other. Configure announce messages (PR #81383), task delivery visibility (PR #81319), and session bootstrapping (PR #81365).
No handoffs defined. Click "Add Handoff" to create a task delivery chain between agents.
Validation
Config looks good! No issues detected.
0
Errors
0
Warnings
0
Info
Live YAML Preview
agents:
defaults:
handoff:
announceHandoff: true # PR #81383 — in-process announce
deliveryVisible: true # PR #81319 — visible task delivery
bootstrapBeforeSend: true # PR #81365 — pre-init sessions
instances:
- name: coordinator
model: anthropic/claude-sonnet-4-6
role: "Orchestrate tasks across agents and manage workflow"
- name: coder
model: anthropic/claude-sonnet-4-6
role: "Write, review, and refactor code"Deploy OpenClaw with DeployClaw
Get multi-agent orchestration with visible handoffs, announce messages, and session bootstrapping out of the box.
Frequently Asked Questions
What is a subagent handoff in OpenClaw?
A subagent handoff is when one agent delegates a task to another agent in a multi-agent setup. The source agent passes context, instructions, and optionally an announce message to the target agent. Handoffs can be conditional (on success, on failure, on keyword) or unconditional (always). This is the core mechanism for multi-agent orchestration in OpenClaw.
What is in-process announce handoff (PR #81383)?
PR #81383 introduced the ability for subagents to announce handoffs within the same process. Previously, handoff announcements required inter-process communication which added latency and complexity. With in-process announce, the source agent can emit a structured handoff message (with template variables like {{from_agent}}, {{to_agent}}, and {{condition}}) that the orchestrator captures and logs — all without leaving the process boundary.
What does visible task delivery mean (PR #81319)?
PR #81319 made subagent task deliveries visible by default. Before this change, when a coordinator handed off a task to a subagent, the delivery was silent — neither the user nor the logging system could see what was being delegated. Now, each task delivery is surfaced with the task description, target agent, and handoff metadata. You can disable visibility per-handoff if you want clean user-facing output.
Why should I enable bootstrap before send (PR #81365)?
PR #81365 introduced session bootstrapping before the first message is sent. Without it, the target agent's session is created lazily on the first message, which can add 200-500ms of cold-start latency. With bootstrapBeforeSend enabled, the session is pre-initialized during the handoff setup phase, so the target agent is ready to process immediately. This is especially important for latency-sensitive workflows.
Can I create circular handoff chains?
Yes, OpenClaw supports circular handoffs (A → B → A) for iterative workflows like code review cycles. However, you should always set a condition (like on_failure or on_keyword) to prevent infinite loops. The session PingPong turn limits (maxPingPongTurns) also act as a safety net. This planner will warn you about potential circular chains.
Does this tool send my config anywhere?
No. The Subagent Handoff Planner runs entirely in your browser. Your agent names, roles, models, and handoff configuration are never sent to any server. The generated YAML is computed locally and can be copied to your clipboard.