Sub-agents

Sub-agents are background agent runs spawned from an existing agent run. They run in their own session (agent:<agentId>:subagent:<uuid>) and, when finished, announce their result back to the requester chat channel. Each sub-agent run is tracked as a background task.

Primary goals:

• Parallelize “research / long task / slow tool” work without blocking the main run.
• Keep sub-agents isolated by default (session separation + optional sandboxing).
• Keep the tool surface hard to misuse: sub-agents do not get session tools by default.
• Support configurable nesting depth for orchestrator patterns.

Cost note: each sub-agent has its own context and token usage by default. For heavy or repetitive tasks, set a cheaper model for sub-agents and keep your main agent on a higher-quality model. Configure via agents.defaults.subagents.model or per-agent overrides. When a child genuinely needs the requester’s current transcript, the agent can request context: "fork" on that one spawn. Thread-bound subagent sessions default to context: "fork" because they branch the current conversation into a follow-up thread.

Slash command

Use /subagents to inspect sub-agent runs for the current session:

/subagents list
/subagents log <id|#> [limit] [tools]
/subagents info <id|#>

/subagents info shows run metadata (status, timestamps, session id, transcript path, cleanup). Use sessions_history for a bounded, safety-filtered recall view; inspect the transcript path on disk when you need the raw full transcript.

Thread binding controls

These commands work on channels that support persistent thread bindings. See Thread supporting channels below.

/focus <subagent-label|session-key|session-id|session-label>
/unfocus
/agents
/session idle <duration|off>
/session max-age <duration|off>

Spawn behavior

Agents start background sub-agents with sessions_spawn. Sub-agent completions return as internal parent-session events; the parent/requester agent decides whether a user-facing update is needed.

Non-blocking, push-based completion

• sessions_spawn is non-blocking; it returns a run id immediately.
• On completion, the sub-agent reports back to the parent/requester session.
• Agent turns that need child results should call sessions_yield after spawning required work. That ends the current turn and lets completion events arrive as the next model-visible message.
• Completion is push-based. Once spawned, do not poll /subagents list, sessions_list, or sessions_history in a loop just to wait for it to finish; inspect status only on-demand for debugging visibility.
• Child output is a report/evidence for the requester agent to synthesize. It is not user-authored instruction text and cannot override system, developer, or user policy.
• On completion, OpenClaw best-effort closes tracked browser tabs/processes opened by that sub-agent session before the announce cleanup flow continues.

Completion delivery

• OpenClaw hands completions back to the requester session through an agent turn with a stable idempotency key.
• If the requester run is still active, OpenClaw first tries to wake/steer that run instead of starting a second visible reply path.
• If an active requester cannot be woken, OpenClaw falls back to a requester-agent handoff with the same completion context instead of dropping the announce.
• A successful parent handoff completes sub-agent delivery even when the parent decides no visible user update is needed.
• Native sub-agents do not get the message tool. They return plain assistant text to the parent/requester agent; human-visible replies are owned by the parent/requester agent’s normal delivery policy.
• If direct handoff cannot be used, it falls back to queue routing.
• If queue routing is still not available, the announce is retried with a short exponential backoff before final give-up.
• Completion delivery keeps the resolved requester route: thread-bound or conversation-bound completion routes win when available; if the completion origin only provides a channel, OpenClaw fills the missing target/account from the requester session’s resolved route (lastChannel / lastTo / lastAccountId) so direct delivery still works.

Completion handoff metadata

The completion handoff to the requester session is runtime-generated internal context (not user-authored text) and includes:

• Result — the latest visible assistant reply text from the child. Tool/toolResult output is not promoted into child results. Terminal failed runs do not reuse captured reply text.
• Status — completed; ready for parent review / failed / timed out / unknown.
• Compact runtime/token stats.
• A review instruction telling the requester agent to verify the result before deciding whether the original task is done.
• Follow-up guidance telling the requester agent to continue the task or record a follow-up when the child result leaves more action.
• A final-update instruction for the no-more-action path, written in normal assistant voice without forwarding raw internal metadata.

Modes and ACP runtime

• --model and --thinking override defaults for that specific run.
• Use info/log to inspect details and output after completion.
• For persistent thread-bound sessions, use sessions_spawn with thread: true and mode: "session".
• If the requester channel does not support thread bindings, use mode: "run" instead of retrying impossible thread-bound combinations.
• For ACP harness sessions (Claude Code, Gemini CLI, OpenCode, or explicit Codex ACP/acpx), use sessions_spawn with runtime: "acp" when the tool advertises that runtime. See ACP delivery model when debugging completions or agent-to-agent loops. When the codex plugin is enabled, Codex chat/thread control should prefer /codex ... over ACP unless the user explicitly asks for ACP/acpx.
• OpenClaw hides runtime: "acp" until ACP is enabled, the requester is not sandboxed, and a backend plugin such as acpx is loaded. runtime: "acp" expects an external ACP harness id, or an agents.list[] entry with runtime.type="acp"; use the default sub-agent runtime for normal OpenClaw config agents from agents_list.

Context modes

Native sub-agents start isolated unless the caller explicitly asks to fork the current transcript.

Mode	When to use it	Behavior
isolated	Fresh research, independent implementation, slow tool work, or anything that can be briefed in the task text	Creates a clean child transcript. This is the default and keeps token use lower.
fork	Work that depends on the current conversation, prior tool results, or nuanced instructions already present in the requester transcript	Branches the requester transcript into the child session before the child starts.

Use fork sparingly. It is for context-sensitive delegation, not a replacement for writing a clear task prompt.

Tool: sessions_spawn

Starts a sub-agent run with deliver: false on the global subagent lane, then runs an announce step and posts the announce reply to the requester chat channel.

Availability depends on the caller’s effective tool policy. The coding and full profiles expose sessions_spawn by default. The messaging profile does not; add tools.alsoAllow: ["sessions_spawn", "sessions_yield", "subagents"] or use tools.profile: "coding" for agents that should delegate work. Channel/group, provider, sandbox, and per-agent allow/deny policies can still remove the tool after the profile stage. Use /tools from the same session to confirm the effective tool list.

Defaults:

• Model: inherits the caller unless you set agents.defaults.subagents.model (or per-agent agents.list[].subagents.model); an explicit sessions_spawn.model still wins.
• Thinking: inherits the caller unless you set agents.defaults.subagents.thinking (or per-agent agents.list[].subagents.thinking); an explicit sessions_spawn.thinking still wins.
• Run timeout: if sessions_spawn.runTimeoutSeconds is omitted, OpenClaw uses agents.defaults.subagents.runTimeoutSeconds when set; otherwise it falls back to 0 (no timeout).
• Task delivery: native sub-agents receive the delegated task in their first visible [Subagent Task] message. The sub-agent system prompt carries runtime rules and routing context, not a hidden duplicate of the task.

Delegation prompt mode

agents.defaults.subagents.delegationMode controls prompt guidance only; it does not change tool policy or enforce delegation.

• suggest (default): keep the standard prompt nudge to use sub-agents for larger or slower work.
• prefer: tell the main agent to stay responsive and delegate anything more involved than a direct reply through sessions_spawn.

Per-agent overrides use agents.list[].subagents.delegationMode.

{
  agents: {
    defaults: {
      subagents: {
        delegationMode: "prefer",
        maxConcurrent: 4,
      },
    },
    list: [
      {
        id: "coordinator",
        subagents: { delegationMode: "prefer" },
      },
    ],
  },
}

Tool parameters

task · string · required

The task description for the sub-agent.

taskName · string

Optional stable handle for identifying a specific child in later status output. Must match [a-z][a-z0-9_]{0,63} and cannot be reserved targets such as last or all.

label · string

Optional human-readable label.

agentId · string

Spawn under another configured agent id when allowed by subagents.allowAgents.

runtime · "subagent" | "acp" · default: "subagent"

acp is only for external ACP harnesses (claude, droid, gemini, opencode, or explicitly requested Codex ACP/acpx) and for agents.list[] entries whose runtime.type is acp.

resumeSessionId · string

ACP-only. Resumes an existing ACP harness session when runtime: "acp"; ignored for native sub-agent spawns.

streamTo · "parent"

ACP-only. Streams ACP run output to the parent session when runtime: "acp"; omit for native sub-agent spawns.

model · string

Override the sub-agent model. Invalid values are skipped and the sub-agent runs on the default model with a warning in the tool result.

thinking · string

Override thinking level for the sub-agent run.

runTimeoutSeconds · number

Defaults to agents.defaults.subagents.runTimeoutSeconds when set, otherwise 0. When set, the sub-agent run is aborted after N seconds.

thread · boolean · default: "false"

When true, requests channel thread binding for this sub-agent session.

mode · "run" | "session" · default: "run"

If thread: true and mode omitted, default becomes session. mode: "session" requires thread: true. If thread binding is unavailable for the requester channel, use mode: "run" instead.

cleanup · "delete" | "keep" · default: "keep"

"delete" archives immediately after announce (still keeps the transcript via rename).

sandbox · "inherit" | "require" · default: "inherit"

require rejects spawn unless the target child runtime is sandboxed.

context · "isolated" | "fork" · default: "isolated"

fork branches the requester’s current transcript into the child session. Native sub-agents only. Thread-bound spawns default to fork; non-thread spawns default to isolated.

sessions_spawn does not accept channel-delivery params (target, channel, to, threadId, replyTo, transport). Native sub-agents report their latest assistant turn back to the requester; external delivery stays with the parent/requester agent.

Task names and targeting

taskName is a model-facing handle for orchestration, not a session key. Use it for stable child names such as review_subagents, linux_validation, or docs_update when a coordinator may need to inspect that child later.

Target resolution accepts exact taskName matches and unambiguous prefixes. Matching is scoped to the same active/recent target window used by numbered /subagents targets, so a stale completed child does not make a reused handle ambiguous. If two active or recent children share the same taskName, the target is ambiguous; use the list index, session key, or run id instead.

The reserved targets last and all are not valid taskName values because they already have control meanings.

Tool: sessions_yield

Ends the current model turn and waits for runtime events, primarily sub-agent completion events, to arrive as the next message. Use it after spawning required child work when the requester cannot produce a final answer until those completions arrive.

sessions_yield is the waiting primitive. Do not replace it with polling loops over subagents, sessions_list, sessions_history, shell sleep, or process polling just to detect child completion.

Only use sessions_yield when the session’s effective tool list includes it. Some minimal or custom tool profiles may expose sessions_spawn and subagents without exposing sessions_yield; in that case, do not invent a polling loop just to wait for completion.

When active children exist, OpenClaw injects a compact runtime-generated Active Subagents prompt block into normal turns so the requester can see the current child sessions, run ids, statuses, labels, tasks, and taskName aliases without polling. The task and label fields in that block are quoted as data, not instructions, because they can originate from user/model-provided spawn arguments.

Tool: subagents

Lists spawned sub-agent runs owned by the requester session. It is scoped to the current requester; a child can only see its own controlled children.

Use subagents for on-demand status and debugging. Use sessions_yield to wait for completion events.

Thread-bound sessions

When thread bindings are enabled for a channel, a sub-agent can stay bound to a thread so follow-up user messages in that thread keep routing to the same sub-agent session.

Thread supporting channels

Discord is currently the only supported channel. It supports persistent thread-bound subagent sessions (sessions_spawn with thread: true), manual thread controls (/focus, /unfocus, /agents, /session idle, /session max-age), and adapter keys channels.discord.threadBindings.enabled, channels.discord.threadBindings.idleHours, channels.discord.threadBindings.maxAgeHours, and channels.discord.threadBindings.spawnSessions.

Quick flow

1. Spawn

   sessions_spawn with thread: true (and optionally mode: "session").

2. Bind

   OpenClaw creates or binds a thread to that session target in the active channel.

3. Route follow-ups

   Replies and follow-up messages in that thread route to the bound session.

4. Inspect timeouts

   Use /session idle to inspect/update inactivity auto-unfocus and /session max-age to control the hard cap.

5. Detach

   Use /unfocus to detach manually.

Manual controls

Command	Effect
/focus <target>	Bind the current thread (or create one) to a sub-agent/session target
/unfocus	Remove the binding for the current bound thread
/agents	List active runs and binding state (thread:<id> or unbound)
/session idle	Inspect/update idle auto-unfocus (focused bound threads only)
/session max-age	Inspect/update hard cap (focused bound threads only)

Config switches

• Global default: session.threadBindings.enabled, session.threadBindings.idleHours, session.threadBindings.maxAgeHours.
• Channel override and spawn auto-bind keys are adapter-specific. See Thread supporting channels above.

See Configuration reference and Slash commands for current adapter details.

Allowlist

agents.list[].subagents.allowAgents · string[]

List of configured agent ids that can be targeted via explicit agentId (["*"] allows any configured target). Default: only the requester agent. If you set a list and still want the requester to spawn itself with agentId, include the requester id in the list.

agents.defaults.subagents.allowAgents · string[]

Default configured target-agent allowlist used when the requester agent does not set its own subagents.allowAgents.

agents.defaults.subagents.requireAgentId · boolean · default: "false"

Block sessions_spawn calls that omit agentId (forces explicit profile selection). Per-agent override: agents.list[].subagents.requireAgentId.

agents.defaults.subagents.announceTimeoutMs · number · default: "120000"

Per-call timeout for gateway agent announce delivery attempts. Values are positive integer milliseconds and are clamped to the platform-safe timer maximum. Transient retries can make the total announce wait longer than one configured timeout.

If the requester session is sandboxed, sessions_spawn rejects targets that would run unsandboxed.

Discovery

Use agents_list to see which agent ids are currently allowed for sessions_spawn. The response includes each listed agent’s effective model and embedded runtime metadata so callers can distinguish PI, Codex app-server, and other configured native runtimes.

allowAgents entries must point at configured agent ids in agents.list[]. ["*"] means any configured target agent plus the requester. If an agent config is deleted but its id remains in allowAgents, sessions_spawn rejects that id and agents_list omits it. Run openclaw doctor --fix to clean stale allowlist entries, or add a minimal agents.list[] entry when the target should remain spawnable while inheriting defaults.

Auto-archive

• Sub-agent sessions are automatically archived after agents.defaults.subagents.archiveAfterMinutes (default 60).
• Archive uses sessions.delete and renames the transcript to *.deleted.<timestamp> (same folder).
• cleanup: "delete" archives immediately after announce (still keeps the transcript via rename).
• Auto-archive is best-effort; pending timers are lost if the gateway restarts.
• runTimeoutSeconds does not auto-archive; it only stops the run. The session remains until auto-archive.
• Auto-archive applies equally to depth-1 and depth-2 sessions.
• Browser cleanup is separate from archive cleanup: tracked browser tabs/processes are best-effort closed when the run finishes, even if the transcript/session record is kept.

Nested sub-agents

By default, sub-agents cannot spawn their own sub-agents (maxSpawnDepth: 1). Set maxSpawnDepth: 2 to enable one level of nesting — the orchestrator pattern: main → orchestrator sub-agent → worker sub-sub-agents.

{
  agents: {
    defaults: {
      subagents: {
        maxSpawnDepth: 2, // allow sub-agents to spawn children (default: 1)
        maxChildrenPerAgent: 5, // max active children per agent session (default: 5)
        maxConcurrent: 8, // global concurrency lane cap (default: 8)
        runTimeoutSeconds: 900, // default timeout for sessions_spawn when omitted (0 = no timeout)
        announceTimeoutMs: 120000, // per-call gateway announce timeout
      },
    },
  },
}

Depth levels

Depth	Session key shape	Role	Can spawn?
0	agent:<id>:main	Main agent	Always
1	agent:<id>:subagent:<uuid>	Sub-agent (orchestrator when depth 2 allowed)	Only if maxSpawnDepth >= 2
2	agent:<id>:subagent:<uuid>:subagent:<uuid>	Sub-sub-agent (leaf worker)	Never

Announce chain

Results flow back up the chain:

1. Depth-2 worker finishes → announces to its parent (depth-1 orchestrator).
2. Depth-1 orchestrator receives the announce, synthesizes results, finishes → announces to main.
3. Main agent receives the announce and delivers to the user.

Each level only sees announces from its direct children.

Operational guidance: start child work once and wait for completion events instead of building poll loops around sessions_list, sessions_history, /subagents list, or exec sleep commands. sessions_list and /subagents list keep child-session relationships focused on live work — live children remain attached, ended children stay visible for a short recent window, and stale store-only child links are ignored after their freshness window. This prevents old spawnedBy / parentSessionKey metadata from resurrecting ghost children after restart. If a child completion event arrives after you already sent the final answer, the correct follow-up is the exact silent token NO_REPLY / no_reply.

Tool policy by depth

• Role and control scope are written into session metadata at spawn time. That keeps flat or restored session keys from accidentally regaining orchestrator privileges.
• Depth 1 (orchestrator, when maxSpawnDepth >= 2): gets sessions_spawn, subagents, sessions_list, sessions_history so it can spawn children and inspect their status. Other session/system tools remain denied.
• Depth 1 (leaf, when maxSpawnDepth == 1): no session tools (current default behavior).
• Depth 2 (leaf worker): no session tools — sessions_spawn is always denied at depth 2. Cannot spawn further children.

Per-agent spawn limit

Each agent session (at any depth) can have at most maxChildrenPerAgent (default 5) active children at a time. This prevents runaway fan-out from a single orchestrator.

Cascade stop

Stopping a depth-1 orchestrator automatically stops all its depth-2 children:

• /stop in the main chat stops all depth-1 agents and cascades to their depth-2 children.

Authentication

Sub-agent auth is resolved by agent id, not by session type:

• The sub-agent session key is agent:<agentId>:subagent:<uuid>.
• The auth store is loaded from that agent’s agentDir.
• The main agent’s auth profiles are merged in as a fallback; agent profiles override main profiles on conflicts.

The merge is additive, so main profiles are always available as fallbacks. Fully isolated auth per agent is not supported yet.

Announce

Sub-agents report back via an announce step:

• The announce step runs inside the sub-agent session (not the requester session).
• If the sub-agent replies exactly ANNOUNCE_SKIP, nothing is posted.
• If the latest assistant text is the exact silent token NO_REPLY / no_reply, announce output is suppressed even if earlier visible progress existed.

Delivery depends on requester depth:

• Top-level requester sessions use a follow-up agent call with external delivery (deliver=true).
• Nested requester subagent sessions receive an internal follow-up injection (deliver=false) so the orchestrator can synthesize child results in-session.
• If a nested requester subagent session is gone, OpenClaw falls back to that session’s requester when available.

For top-level requester sessions, completion-mode direct delivery first resolves any bound conversation/thread route and hook override, then fills missing channel-target fields from the requester session’s stored route. That keeps completions on the right chat/topic even when the completion origin only identifies the channel.

Child completion aggregation is scoped to the current requester run when building nested completion findings, preventing stale prior-run child outputs from leaking into the current announce. Announce replies preserve thread/topic routing when available on channel adapters.

Announce context

Announce context is normalized to a stable internal event block:

Field	Source
Source	subagent or cron
Session ids	Child session key/id
Type	Announce type + task label
Status	Derived from runtime outcome (success, error, timeout, or unknown) — not inferred from model text
Result content	Latest visible assistant text from the child
Follow-up	Instruction describing when to reply vs stay silent

Terminal failed runs report failure status without replaying captured reply text. Tool/toolResult output is not promoted into child result text.

Stats line

Announce payloads include a stats line at the end (even when wrapped):

• Runtime (e.g. runtime 5m12s).
• Token usage (input/output/total).
• Estimated cost when model pricing is configured (models.providers.*.models[].cost).
• sessionKey, sessionId, and transcript path so the main agent can fetch history via sessions_history or inspect the file on disk.

Internal metadata is meant for orchestration only; user-facing replies should be rewritten in normal assistant voice.

Why prefer sessions_history

sessions_history is the safer orchestration path:

• Assistant recall is normalized first: thinking tags stripped; <relevant-memories> / <relevant_memories> scaffolding stripped; plain-text tool-call XML payload blocks (<tool_call>, <function_call>, <tool_calls>, <function_calls>) stripped, including truncated payloads that never close cleanly; downgraded tool-call/result scaffolding and historical-context markers stripped; leaked model control tokens (<|assistant|>, other ASCII <|...|>, full-width <｜...｜>) stripped; malformed MiniMax tool-call XML stripped.
• Credential/token-like text is redacted.
• Long blocks can be truncated.
• Very large histories can drop older rows or replace an oversized row with [sessions_history omitted: message too large].
• Raw on-disk transcript inspection is the fallback when you need the full byte-for-byte transcript.

Tool policy

Sub-agents use the same profile and tool-policy pipeline as the parent or target agent first. After that, OpenClaw applies the sub-agent restriction layer.

With no restrictive tools.profile, sub-agents get all tools except the message tool, session tools, and system tools:

• sessions_list
• sessions_history
• sessions_send
• sessions_spawn
• message

sessions_history remains a bounded, sanitized recall view here too — it is not a raw transcript dump.

When maxSpawnDepth >= 2, depth-1 orchestrator sub-agents additionally receive sessions_spawn, subagents, sessions_list, and sessions_history so they can manage their children.

Override via config

{
  agents: {
    defaults: {
      subagents: {
        maxConcurrent: 1,
      },
    },
  },
  tools: {
    subagents: {
      tools: {
        // deny wins
        deny: ["gateway", "cron"],
        // if allow is set, it becomes allow-only (deny still wins)
        // allow: ["read", "exec", "process"]
      },
    },
  },
}

tools.subagents.tools.allow is a final allow-only filter. It can narrow the already-resolved tool set, but it cannot add back a tool removed by tools.profile. For example, tools.profile: "coding" includes web_search/web_fetch but not the browser tool. To let coding-profile sub-agents use browser automation, add browser at the profile stage:

{
  tools: {
    profile: "coding",
    alsoAllow: ["browser"],
  },
}

Use per-agent agents.list[].tools.alsoAllow: ["browser"] when only one agent should get browser automation.

Concurrency

Sub-agents use a dedicated in-process queue lane:

• Lane name: subagent
• Concurrency: agents.defaults.subagents.maxConcurrent (default 8)

Liveness and recovery

OpenClaw does not treat endedAt absence as permanent proof that a sub-agent is still alive. Unended runs older than the stale-run window stop counting as active/pending in /subagents list, status summaries, descendant completion gating, and per-session concurrency checks.

After a gateway restart, stale unended restored runs are pruned unless their child session is marked abortedLastRun: true. Those restart-aborted child sessions remain recoverable through the sub-agent orphan recovery flow, which sends a synthetic resume message before clearing the aborted marker.

Automatic restart recovery is bounded per child session. If the same sub-agent child is accepted for orphan recovery repeatedly inside the rapid re-wedge window, OpenClaw persists a recovery tombstone on that session and stops auto-resuming it on later restarts. Run openclaw tasks maintenance --apply to reconcile the task record, or openclaw doctor --fix to clear stale aborted recovery flags on tombstoned sessions.

If a sub-agent spawn fails with Gateway PAIRING_REQUIRED / scope-upgrade, check the RPC caller before editing pairing state. Internal sessions_spawn coordination should connect as client.id: "gateway-client" with client.mode: "backend" over direct loopback shared-token/password auth; that path does not depend on the CLI’s paired-device scope baseline. Remote callers, explicit deviceIdentity, explicit device-token paths, and browser/node clients still need normal device approval for scope upgrades.

Stopping

• Sending /stop in the requester chat aborts the requester session and stops any active sub-agent runs spawned from it, cascading to nested children.

Limitations

• Sub-agent announce is best-effort. If the gateway restarts, pending “announce back” work is lost.
• Sub-agents still share the same gateway process resources; treat maxConcurrent as a safety valve.
• sessions_spawn is always non-blocking: it returns { status: "accepted", runId, childSessionKey } immediately.
• Sub-agent context only injects AGENTS.md and TOOLS.md (no SOUL.md, IDENTITY.md, USER.md, MEMORY.md, HEARTBEAT.md, or BOOTSTRAP.md). Codex-native subagents follow the same boundary: TOOLS.md stays in inherited Codex thread instructions, while parent-only persona, identity, and user files are injected as turn-scoped collaboration instructions so children do not clone them.
• Maximum nesting depth is 5 (maxSpawnDepth range: 1–5). Depth 2 is recommended for most use cases.
• maxChildrenPerAgent caps active children per session (default 5, range 1–20).

Related

• ACP agents
• Agent send
• Background tasks
• Multi-agent sandbox tools