openclaw approvals

Manage exec approvals for the local host, gateway host, or a node host. With no target flag, commands read/write the local approvals document in shared SQLite state. Use --gateway to target the gateway, or --node <id|name|ip> to target a specific node.

Alias: openclaw exec-approvals

Related: Exec approvals, Nodes

Common commands

openclaw approvals getopenclaw approvals get --node <id|name|ip>openclaw approvals get --gatewayopenclaw approvals pendingopenclaw approvals resolve <id> <allow-once|allow-always|deny>

get shows the effective exec policy for the target: the requested tools.exec policy, the host approvals-file policy, and the merged effective result. Nodes with a host-native policy, such as the Windows companion, show that policy directly instead of applying OpenClaw approvals-file policy math.

For file-backed nodes, the merged view requires a host-resolved policy snapshot. Older nodes show the effective policy as unavailable instead of assuming the Gateway's requested policy also applies on the host.

Precedence:

• The host approvals document is the enforceable source of truth.
• Requested tools.exec policy can narrow or broaden intent, but the effective result is derived from host rules.
• --node combines the node host approvals document with gateway tools.exec policy (both apply at runtime).
• If gateway config is unavailable, the CLI falls back to the node approvals snapshot and notes that the final runtime policy could not be computed.

Pending approvals

List pending exec, plugin, and OpenClaw system-agent approvals from the Gateway:

openclaw approvals pendingopenclaw approvals pending --json

Complete enumeration and the matching operator-wide resolve flow use operator.admin because approval records otherwise retain requester/reviewer filtering. Resolution also requests the dedicated operator.approvals scope. The standard CLI operator grant includes both scopes; a restricted third-party client should not request admin merely to emulate this command.

Human output shows the approval kind, agent/session attribution, request age, time until expiry, a shortened command or summary, and a shell-neutral id64_<base64url> id token. A Full request text block always follows the compact table with every complete token and a losslessly escaped request, so terminal-width shortening cannot hide a suffix or the token needed for resolution. Copy the complete token into resolve. Unsafe terminal characters in other fields are shown as visible Unicode escapes. JSON output returns normalized entries under approvals, preserving the original raw id, summary, createdAtMs, and expiresAtMs for scripts; raw ids remain accepted by resolve unless they use the reserved id64_ display-token prefix.

If a supplied id64_ value matches both a literal raw id and the decoded display token for another approval, the CLI rejects it as ambiguous instead of risking resolution of the wrong request.

Resolve one approval by its full id:

openclaw approvals resolve <id> allow-onceopenclaw approvals resolve <id> allow-alwaysopenclaw approvals resolve <id> deny --reason "Not expected during maintenance"

For exec requests, allow-always means always allow here: the generated grant is tied to the command's exact arguments and current working directory. The same command from another directory requires a separate approval.

For approvals raised by an automation (cron) run, allow-always mints a scoped standing grant instead of a JSON allowlist entry (see Standing grants for automations). By default the grant lives until revoked; --expires-in-days <n> freezes an explicit lifetime instead of the configured tools.exec.grantExpiryDays default:

openclaw approvals resolve <id> allow-always --expires-in-days 30

Standing grants

Standing grants minted by allow-always on automation approvals are listed and revoked from the same command group:

openclaw approvals grants listopenclaw approvals grants list --jsonopenclaw approvals grants revoke <grant-id>

The list shows the owning automation, the exact command, the use count, and each grant's state (until revoked, expires in N days, expired, or revoked). Revocation is idempotent and takes effect at the next occurrence's spawn boundary — that occurrence prompts again. Editing or deleting the automation invalidates its grants without needing an explicit revoke.

The CLI reads the unified approval record to select its kind, checks the requested decision against the record's allowed decisions, and then calls the unified resolver. A first successful decision exits 0. Repeating the recorded decision also exits 0 and reports already resolved (same decision). A conflicting decision, missing approval, expired approval, or decision unavailable for that approval kind prints a clear error and exits non-zero.

--reason adds a local note to the CLI confirmation. The current Gateway approval record has no free-text resolution-reason field, so this note is not persisted or sent to other approval surfaces.

Replace approvals from a file

openclaw approvals set --file ./exec-approvals.jsonopenclaw approvals set --stdin <<'EOF'{ version: 1, defaults: { security: "full", ask: "off", askFallback: "full" } }EOFopenclaw approvals set --node <id|name|ip> --file ./exec-approvals.jsonopenclaw approvals set --gateway --file ./exec-approvals.json

set accepts JSON5, not only strict JSON. Use either --file or --stdin, not both.

Host-native Windows nodes use their own policy shape:

openclaw approvals set --node <id|name|ip> --stdin <<'EOF'{  defaultAction: "deny",  rules: [{ pattern: "hostname", action: "allow" }]}EOF

The CLI reads the node's current hash first and sends it with the update, so concurrent local edits are rejected instead of overwritten. rules is required because this operation replaces the node's complete rule list; defaultAction is optional. A node that reports its native policy as disabled cannot be configured remotely; enable or configure the policy on that host first. Host-native policies do not support the allowlist add|remove helpers.

"Never prompt" / YOLO example

Set the host approvals defaults to full + off for a host that should never stop on exec approvals:

openclaw approvals set --stdin <<'EOF'{  version: 1,  defaults: {    security: "full",    ask: "off",    askFallback: "full"  }}EOF

For nodes that expose an OpenClaw approvals document, use the same body with openclaw approvals set --node <id|name|ip> --stdin. Host-native nodes require their owner-specific shape shown above.

This changes the host approvals document only. To keep the requested OpenClaw policy aligned, also set:

openclaw config set tools.exec.host gatewayopenclaw config set tools.exec.mode full

tools.exec.host=gateway is explicit here because host=auto still means "sandbox when available, otherwise gateway": YOLO is about approvals, not routing. Use gateway (or /exec host=gateway) when you want host exec even with a sandbox configured.

Omitted askFallback defaults to deny. Set askFallback: "full" explicitly when upgrading a no-UI host that should keep never-prompt behavior.

Local shortcut for the same intent, on the local machine only:

openclaw exec-policy preset yolo

Allowlist helpers

openclaw approvals allowlist add "~/path/to/**/bin/rg"openclaw approvals allowlist add --agent main --node <id|name|ip> "/usr/bin/uptime"openclaw approvals allowlist add --agent "*" "/usr/bin/uname" openclaw approvals allowlist remove "~/path/to/**/bin/rg"

Adding an existing pattern or removing a missing one succeeds without writing. With --json, these commands return the unchanged, redacted approvals snapshot.

Common options

get, set, and allowlist add|remove all support:

• --node <id|name|ip> (resolves id, name, IP, or id prefix; same resolver as openclaw nodes)
• --gateway
• shared node RPC options: --url, --token, --timeout, --json

No target flag means the local approvals row in the shared state database.

allowlist add|remove also supports --agent <id> (defaults to "*", applying to all agents).

pending and resolve always use the Gateway because pending requests are live Gateway state. They support the shared Gateway connection options --url, --token, and --timeout; pending also supports --json.

openclaw exec-policy

openclaw exec-policy show explains an agent's terminal tool policies and shows command approvals separately. Inspection never changes permissions. preset and set synchronize requested tools.exec.* config with the local host approvals document.

openclaw exec-policy showopenclaw exec-policy show --agent mainopenclaw exec-policy show --session agent:main:mainopenclaw exec-policy show --agent main --verboseopenclaw exec-policy show --agent main --json openclaw exec-policy preset yoloopenclaw exec-policy preset cautious --json openclaw exec-policy set --host gateway --security full --ask off --ask-fallback full --json

Inspect terminal access

Without --session, show works offline. It reports definitive exclusions from local tool profiles and policies. A locally allowed tool's execution access is unverified; model, channel, sandbox, session, and runtime restrictions may still affect availability.

The report leads with the agent and terminal-access status, followed by profile inheritance, findings for exec and process, local command approvals, and a next step. An agent profile overrides the global profile. For example, agents.entries.main.tools.profile: "messaging" excludes terminal tools even if tools.profile is "full". Command approval settings do not grant tool access. Intentionally restrictive profiles are valid configuration.

• --agent <id> selects the agent to inspect. Without --session, the agent must be configured locally. With a single configured agent, selection is automatic. With multiple agents, an untargeted report asks you to select an agent or an agent-qualified session key before inspecting tool availability. With no configured agents, it explains how to add one. Both cases still show a compact summary of all local command approval scopes.
• --session <key> fetches a read-only tool preview for an existing session from its saved settings through the Gateway. Use the full session key; for a shared key such as global, the Gateway resolves its agent; pass --agent when the Gateway needs an explicit selection. The agent need not exist in the CLI machine's configuration. Conflicting agent and session targets are rejected.
• --verbose adds policy sources and the complete requested/host/effective approval tables for all scopes.
• --json preserves all approval fields and scopes. With a selected agent, it adds toolAccess, separating local findings from the optional live result. local is omitted when the agent is not configured on the CLI machine. If a failed session inspection could not resolve an agent, agentId is also omitted. An untargeted multi-agent or empty-roster report instead adds toolAccessSelectionRequired with agentIds and a hint; it makes no tool availability claim.

The session preview supports the shared Gateway connection options (--url, --port, --token, --password, --timeout). If the Gateway or session cannot be inspected, the report says UNVERIFIED and retains any available local findings. Missing local tool policy is labeled unavailable, never allowed. An unavailable inspection is not evidence that tools are allowed or denied.

A successful fetch reports PREVIEW unless the checked policies establish an exclusion. Included tools are not guaranteed to execute, and a tool missing from the preview is not necessarily disabled. The preview may be cached; saved changes can take time to appear, and an active run can use different authority, credentials, discovery, or policy. Verify execution in a run; command approvals still apply.

JSON retains the wire names live, checked: "live-session", and tool statuses available/unavailable. These describe the fetched session preview. live.status: "verified" means the preview was retrieved successfully, not that execution was verified. excluded identifies a checked policy exclusion; unavailable alone does not establish one. The CLI translates an older Gateway's explicit deniedBySession flag to excluded with a session reason.

Command approvals remain labeled local, including during session inspection. They do not describe remote host approvals or per-session /exec overrides. Use openclaw approvals get --gateway or --node <id|name|ip> for those host policies, and /exec in the session for its current defaults.

When adding terminal tools to alsoAllow would address the checked restrictions, the report suggests the exact configuration path and tool names. Append them to existing entries; preserve command approval requirements. When another policy still blocks access, review the reported exclusions before enabling tools. A tool merely missing from the preview does not prompt an alsoAllow recipe. Changes use the existing config commands or the agent's Tools panel and Save controls. Check the updated preview and verify execution in a run.

Synchronize local command approvals

Presets (yolo, cautious, deny-all) apply host, security, ask, and askFallback together. set applies only the flags you pass; each accepted value is validated (--host auto|sandbox|gateway|node, --security deny|allowlist|full, --ask off|on-miss|always, --ask-fallback deny|allowlist|full).

show, preset, and set accept --json and return the requested, host, and effective command approval facts as one JSON object. preset and set do not change tool profiles or tool allow/deny rules.

Scope:

• Updates the local config file and local approvals document together; does not push policy to the gateway or a node host.
• --host node is rejected: node exec approvals are fetched from the node at runtime, so local exec-policy cannot synchronize them. Use openclaw approvals set --node <id|name|ip> instead.
• exec-policy show marks host=node scopes as node-managed at runtime instead of deriving an effective policy from the local approvals document.

For remote host approvals, use openclaw approvals set --gateway or openclaw approvals set --node <id|name|ip> directly.

Notes

• The node host must advertise system.execApprovals.get/set (macOS app, headless node host, or Windows companion).
• Generated grants became directory-bound in 2026.8.1. After upgrading from 2026.7.1 or earlier, run openclaw doctor --fix if the update did not already do so. Doctor removes only inactive generated grants; manual allowlist rules stay in place. Rerun affected workflows to approve them in the intended directory.
• Approvals are stored per host in $OPENCLAW_STATE_DIR/state/openclaw.sqlite#exec_approvals_config, or ~/.openclaw/state/openclaw.sqlite#exec_approvals_config when the variable is unset. The suffix identifies the singleton SQLite row.

Related

• CLI reference
• Exec approvals