Skip to content

NextClaw CLI: Capability Map and Complete Command Reference

The nextclaw CLI is a first-class way to operate NextClaw. When a product capability can be expressed clearly as a structured operation, NextClaw aims to make it available from the command line so users, developers, scripts, CI jobs, and other Agents can work with the same underlying capabilities.

This does not mean moving every interface into a terminal. Visual, drag-and-drop, and direct-manipulation work can remain GUI-first. The CLI is most useful for runtime management, queries and diagnostics, repeatable operations, automation, and system integration.

Use this page to:

  • see what NextClaw can currently do, grouped by capability area; and
  • look up the complete CLI surface by command path.

The registered command tree is the source of truth for this page. Adding, removing, or renaming a CLI command requires updating both language versions, and repository tests verify that this reference covers the real command surface.

How to use this reference

This page lists every executable command path and its primary purpose. To inspect the complete arguments, defaults, and options for one command, run:

bash
nextclaw <command> --help

Use nextclaw --version to inspect the installed version. Many query and management commands support --json; check the command help before integrating it into a script, CI job, or Agent. For first-time setup, start with Quickstart. For the small set of commands used most often, see Core Commands.

Feedback and private discussions

No GitHub login is required; see problem feedback. feedback names the feedback application. Generic threads, posts, participants, and event subscriptions use discussion.

CommandPurpose
nextclaw feedback submitSubmit a private report and save its receipt
nextclaw feedback listList local receipts
nextclaw feedback getRead report details and replies
nextclaw feedback replyAdd reproduction information
nextclaw feedback withdrawWithdraw a report
nextclaw feedback linkLink the current NextClaw account
nextclaw feedback syncSync account reports
nextclaw feedback exportExport a private receipt
nextclaw feedback importRestore a private receipt
nextclaw feedback workflow skill-pathLocate the packaged feedback workflow skill
nextclaw feedback workflow listRead the approved feedback work queue
nextclaw feedback workflow getRead a report and its current approval
nextclaw feedback workflow claimClaim approved feedback
nextclaw feedback workflow commentWrite a participant reply
nextclaw feedback workflow resultSubmit verification evidence or a blocker
nextclaw feedback workflow triageClassify without granting repair permission
nextclaw feedback workflow recoverRecover after confirming the previous run stopped
nextclaw feedback workflow authorize-deliveryAssociate an approved repair commit
nextclaw feedback workflow publishVerify release proof and update the report
nextclaw discussion skill-pathLocate the packaged discussion participant skill
nextclaw discussion listList accessible threads by space
nextclaw discussion eventsRead events addressed to the current role after a cursor
nextclaw discussion getRead a thread and its ordered posts
nextclaw discussion postPost as the authenticated participant
nextclaw discussion listen configureSave the endpoint, participant credential, and consumer command; the Codex preset can take a workspace
nextclaw discussion listen startStart the code-only listener and complete its first scan
nextclaw discussion listen statusInspect listener health, heartbeat, and the latest error
nextclaw discussion listen stopStop the current listener instance
nextclaw discussion listen restartRestart the listener with its saved configuration
nextclaw discussion listen workerInternal listener process entry point
nextclaw discussion listen codex-desktop-triggerInternal Codex Desktop consumer preset entry point
nextclaw discussion listen codex-desktop-runnerInternal asynchronous Codex Desktop task runner

Capability map

AreaWhat you can do
Setup and accountInitialize a workspace, sign in, inspect the account, and set a username
Runtime, status, and logsStart and stop the local service and UI, update, diagnose, inspect logs, and view model usage
Host managementInstall or remove autostart services on Linux, macOS, and Windows
Remote accessEnable, disable, diagnose, and debug remote connections
Agents and task executionChat in a terminal, run Headless tasks, and manage Agents and Runtimes
Projects and sessionsCreate projects, manage work items and artifact links, and organize session bindings
Automation and learning loopManage scheduled jobs and learning-loop policy
Configuration and secretsRead and write configuration, audit secrets, and apply secret references
MCP and messaging channelsManage MCP servers and messaging-channel connections
Skills and MarketplaceInspect, install, publish, update, and discover Skills
NextClaw AppsCheck, develop, package, publish, install, call, and manage App data

Setup and account

CommandPurpose
nextclaw onboardInitialize NextClaw configuration and a workspace
nextclaw initInitialize a workspace; use --force to overwrite template files
nextclaw loginSign in to NextClaw Platform and save local credentials
nextclaw account statusInspect account status and Marketplace publishing readiness
nextclaw account set-usernameSet the username used for personal Marketplace publishing

Runtime, status, and logs

CommandPurpose
nextclaw gatewayStart the Gateway in the foreground, optionally with the UI
nextclaw uiStart the Gateway and UI
nextclaw startStart the Gateway and UI in the background
nextclaw restartRestart the service; supported hosts resume sessions interrupted by this restart
nextclaw serveRun the Gateway and UI in the foreground for debugging
nextclaw stopStop the background service
nextclaw statusInspect the current installation, storage paths, running runtime version, processes, health, configuration, and endpoints
nextclaw doctorRun diagnostics and optionally repair safe stale state
nextclaw logs pathShow local log-file paths
nextclaw logs tailShow recent service or crash logs
nextclaw logs queryQuery structured logs by time, level, domain, event, or correlation ID
nextclaw usageInspect recent model usage, history, and cache statistics
nextclaw updateCheck, download, or apply a NextClaw Runtime update

With explicit user authorization, an agent can run ordinary nextclaw update, then nextclaw restart if required. Managed services, foreground hosts with the local runtime API, and NextClaw systemd services record active sessions before exit. Once ready, the replacement process continues both the initiating session and concurrent sessions once, with context stating that the restart succeeded and interrupted commands must not be replayed.

Use plain nextclaw restart, without port/open/timeout overrides, for this handoff. Desktop and other-supervisor exits, legacy stop/start, crashes, and ordinary startup are outside its scope. A request timeout does not prove failure: check nextclaw status --json before retrying. Its runtime.version field is reported by the process serving the local API and is authoritative for the running host; nextclaw --version describes the CLI/runtime selected for that command. Completed or ordinarily failed tasks are not automatically rerun.

For the current instance's data locations, use instance and storage from nextclaw status --json. In Portable Edition, storage.portableDataRoot is <portableRoot>/data and storage.runtimeHome is its runtime-home child. Desktop logs and Runtime structured logs are reported separately as desktopLogsDirectory and runtimeLogsDirectory. .nextclaw is only the ordinary default when no override or Portable Edition profile applies; it is not a substitute for querying the current instance.

Host management and autostart

For the user workflow, restart-continuation behavior, and complete limits, see Self-management.

CommandPurpose
nextclaw service install-systemdInstall a Linux user-level or system-level systemd service
nextclaw service uninstall-systemdRemove a NextClaw-managed systemd service
nextclaw service install-launch-agentInstall the macOS LaunchAgent
nextclaw service uninstall-launch-agentRemove the macOS LaunchAgent
nextclaw service install-taskInstall the Windows Scheduled Task
nextclaw service uninstall-taskRemove the Windows Scheduled Task
nextclaw service autostart statusInspect the host autostart owner and state
nextclaw service autostart doctorDiagnose host autostart configuration

Remote access

CommandPurpose
nextclaw remote enableEnable service-managed remote access
nextclaw remote disableDisable remote access
nextclaw remote statusInspect remote-access and connection state
nextclaw remote doctorRun remote-access diagnostics
nextclaw remote connectRegister the device and keep it connected in foreground debug mode

Agents and task execution

CommandPurpose
nextclaw agentChat in the terminal or send one message with -m
nextclaw execRun one Headless task with text, JSON, or JSONL output; see nextclaw exec
nextclaw agents listList configured Agents
nextclaw agents runtimesList and optionally probe available Agent Runtimes
nextclaw agents runtime configInspect or change configuration for one Runtime
nextclaw agents newCreate an Agent with a name, avatar, home directory, and Runtime
nextclaw agents updateUpdate an existing Agent
nextclaw agents removeRemove an Agent

Projects and sessions

CommandPurpose
nextclaw projects listList registered projects, including projects without sessions
nextclaw projects templatesList built-in project templates
nextclaw projects createCreate and register a project
nextclaw projects registerRegister an existing directory without modifying its contents; accepts optional --name
nextclaw projects removeRemove a project from the list while preserving its folder, sessions, and work; requires exact ID confirmation
nextclaw projects work listPage through work items by project ID with --state, --cursor, --limit, and --include-deleted
nextclaw projects work getShow work-item details
nextclaw projects work createCreate a persistent work item
nextclaw projects work updateUpdate fields, state, or attention
nextclaw projects work deleteSoft-delete a work item
nextclaw projects work restoreRestore a deleted work item
nextclaw projects work activityShow immutable work-item activity
nextclaw projects work artifact linkLink an artifact file inside the project
nextclaw projects work artifact unlinkRemove an artifact link
nextclaw projects work state listList custom project work states
nextclaw projects work state createCreate a work state
nextclaw projects work state updateUpdate or reorder a work state
nextclaw projects work state deleteDelete a state and optionally migrate existing items
nextclaw sessions renameRename a session
nextclaw sessions set-projectBind a session to an existing project directory
nextclaw sessions clear-projectClear a session's explicit project binding
nextclaw sessions deletePermanently delete a session; requires --confirm <session-id>

Every projects work command requires --project <project-id> and runs through the local NextClaw service.

Automation and learning loop

CommandPurpose
nextclaw cron listList scheduled jobs
nextclaw cron addAdd a job using an interval, cron expression, or one-time date
nextclaw cron removeRemove a scheduled job
nextclaw cron enableEnable a scheduled job
nextclaw cron disableDisable a scheduled job
nextclaw cron runRun a selected job immediately
nextclaw learning-loop statusInspect learning-loop settings
nextclaw learning-loop enableEnable the learning loop
nextclaw learning-loop disableDisable the learning loop
nextclaw learning-loop thresholdSet the tool-call threshold for a learning review

Configuration and secrets

These commands require a running NextClaw host and return JSON. Discover templates with providers templates, then create a provider with providers add office --type openai --api-key-env MY_PROVIDER_KEY. Keys are read from the named environment variable; API keys and custom header values are not printed.

bash
nextclaw providers models discover office --json
nextclaw providers models set office <provider-scoped-model-id> --json
nextclaw providers test office --model <provider-scoped-model-id> --json
nextclaw models set <provider-scoped-model-id> --json
nextclaw models show --json
nextclaw providers show office --json

Discovery returns upstream names. Prefix the selected name with the instance id: gpt-example on office becomes office/gpt-example. Preserve upstream slashes and do not duplicate an existing instance prefix. Discovery does not save models. providers models set replaces the entire configured list; omit models to clear it. providers models configure replaces all capability overrides: use repeatable --vision model=true|false, --thinking model=off,high, --thinking-default model=high, or --clear. Query first and include entries to retain. The default thinking level must occur in the supplied supported levels. Vision=false removes the positive override, rather than forcing built-in vision capability off.

Failed connection tests exit non-zero. Start authorization with auth start, follow the returned URI/code, and use auth poll at the returned interval; pending is not completion. Writes await the host's apply step. An apply failure needs investigation even if values were saved. Commands do not restart the host.

For search, run nextclaw search provider exa --api-key-env MY_EXA_KEY, then nextclaw search configure --provider exa --enabled-provider exa --max-results 10 and verify with search show. Result counts accept 1–50. The enabled list is replaced; --clear-enabled-providers disables all. Bocha supports summary/freshness/docs-url; Tavily supports search-depth/include-answer. Other providers reject these specific options. See command help for all flags.

CommandPurpose
nextclaw providers listList provider instances
nextclaw providers templatesList templates and authorization methods
nextclaw providers showInspect one provider
nextclaw providers addAdd a template or custom provider
nextclaw providers updateUpdate name, endpoint, credentials, protocol and headers
nextclaw providers removeRemove a provider and its secret references
nextclaw providers enableEnable a provider
nextclaw providers disableDisable a provider
nextclaw providers testTest a model connection
nextclaw providers models listInspect configured models and capabilities
nextclaw providers models discoverDiscover models without saving
nextclaw providers models setReplace or clear the model list
nextclaw providers models configureReplace or clear model capability overrides
nextclaw providers auth startStart provider authorization
nextclaw providers auth pollPoll authorization once
nextclaw providers auth importImport supported provider CLI credentials
nextclaw models listInspect the runtime model catalog
nextclaw models showInspect the default model
nextclaw models setSet the default model
nextclaw search showInspect search settings
nextclaw search configureSelect default/enabled providers and result count
nextclaw search providerConfigure one search provider

Generic configuration and secrets

Prefer object-level commands for covered tasks. Generic config commands remain for settings without a dedicated command or explicit manual recovery.

CommandPurpose
nextclaw config getRead a configuration value by dot path
nextclaw config setWrite a configuration value by dot path
nextclaw config unsetRemove a configuration value
nextclaw secrets auditAudit secret-reference resolution
nextclaw secrets configureConfigure an env, file, or exec secret provider
nextclaw secrets applyApply secret references and provider configuration in bulk or one at a time
nextclaw secrets reloadTell the running service to reload secrets

MCP and messaging channels

CommandPurpose
nextclaw mcp listList configured MCP servers
nextclaw mcp addAdd a stdio, HTTP, or SSE MCP server
nextclaw mcp removeRemove an MCP server
nextclaw mcp enableEnable an MCP server
nextclaw mcp disableDisable an MCP server
nextclaw mcp doctorCheck MCP connectivity and tool discovery
nextclaw channels addAdd or update messaging-channel configuration
nextclaw channels listList configured channels
nextclaw channels statusInspect channel state
nextclaw channels loginLink a supported channel account by QR code

Skills and Marketplace

| --------------------------------------- | -------------------------------------------- | | nextclaw skills installed | List Skills installed in the current runtime | | nextclaw skills info | Inspect an installed Skill | | nextclaw skills install | Install a Skill from NextClaw Marketplace | | nextclaw skills publish | Create or publish a Marketplace Skill | | nextclaw skills update | Update a published Marketplace Skill | | nextclaw marketplace skills search | Search Marketplace Skills | | nextclaw marketplace skills info | Inspect a Marketplace Skill | | nextclaw marketplace skills recommend | List recommended Skills | | nextclaw marketplace skills install | Install a Marketplace Skill | | nextclaw marketplace skills update | Update a locally installed Marketplace Skill |

NextClaw Apps

CommandPurpose
nextclaw app createCreate a standalone App; the default template is Rust/WASI
nextclaw app doctorDiagnose the WASI Guest build environment and print repair commands
nextclaw app buildBuild Rust/WASI Service Components in an App
nextclaw app checkCheck a complete App package, Panel, or Service directory
nextclaw app testRun an App's Action smoke tests in the isolated Runtime
nextclaw app devStart the real Runtime from an App package or Service directory
nextclaw app packPackage a .napp; pure WASI Apps default to a universal artifact
nextclaw app validate-publishValidate an App and its artifacts before Marketplace submission
nextclaw app publishSubmit an App to App Marketplace
nextclaw app callCall a real Action from an App package or Service directory
nextclaw app restartRestart a Service App running in the NextClaw UI
nextclaw app data listList active and retained App data instances
nextclaw app data deletePermanently delete retained App data with an exact App-ID confirmation
nextclaw app marketplace searchSearch Apps in the official App Marketplace
nextclaw app marketplace infoShow a Marketplace App and its derived install command
nextclaw app listList Apps installed in the running NextClaw host
nextclaw app infoShow installed App state and versions
nextclaw app invokeCall an Action on an enabled installed App through the running host
nextclaw app verificationRead redacted runtime verification records from the running host
nextclaw app acceptance contractRead the stable Portable Runtime acceptance contract
nextclaw app acceptance statusRead current Portable Runtime acceptance status and evidence freshness
nextclaw app acceptance exportExport the contract, current runtime identity, and acceptance status as JSON
nextclaw app jobs listList durable Jobs for one installed App instance
nextclaw app jobs inspectInspect one durable App Job
nextclaw app jobs watchReplay retained Job progress and output after an optional sequence cursor
nextclaw app jobs cancelRequest Job cancellation; completion remains pending until runtime confirmation
nextclaw app resident-inbox listInspect durable Resident delivery state; --dead-letters narrows to recoverable failures
nextclaw app resident-inbox replayReplay one dead-letter Resident event through the host-owned inbox
nextclaw app dependencies inspectInspect external capability/resource dependencies, Provider candidates, and bindings
nextclaw app dependencies verifyVerify whether current dependencies are satisfied
nextclaw app dependencies setupEstablish bindings only when a compatible Provider is unique
nextclaw app dependencies bindBind one dependency to an installed trusted Provider
nextclaw app dependencies unbindRemove one dependency binding
nextclaw app secrets inspectShow declared Secret slots and non-sensitive SecretRef bindings
nextclaw app secrets verifyResolve bindings without revealing Secret values
nextclaw app secrets bindBind one declared Secret slot to an env, file, or exec provider
nextclaw app secrets unbindRemove an App SecretRef binding and its active Secret permission
nextclaw app permissions inspectInspect declared directory scopes, grant status, and effective access mode
nextclaw app permissions document grantGrant or replace a runtime-host directory as read-only or read-write
nextclaw app permissions document revokeRevoke a directory scope and stop the previous mount
nextclaw app ai-capabilities inspectInspect declared non-secret model and Agent slots with current bindings
nextclaw app ai-capabilities verifyVerify required model and Agent slot readiness
nextclaw app ai-capabilities bindBind one declared model or Agent slot to a configured target
nextclaw app ai-capabilities unbindRemove one model or Agent slot binding
nextclaw app operationsList durable App lifecycle operations
nextclaw app installInstall a Marketplace App, local directory, or .napp bundle through the running host
nextclaw app enableEnable an installed App
nextclaw app disableDisable an installed App
nextclaw app updateStart a background App update
nextclaw app rollbackRoll back to an installed version
nextclaw app uninstallStart an uninstall; purging data requires exact App-ID confirmation

See Service Apps for the user workflow and Develop a WASM Service App for runtime development commands.

app dev and app call accept a schema v2 App root directly. A package with one Service is selected automatically; use --component <service-id> when a package has multiple Services. Local .napp files can be installed by relative path, for example nextclaw app install ./my-app.napp.

app list --json exposes each App's primaryPanelId and components[].id. The built-in show_panel_app tool accepts either an App ID, which opens its enabled primary Panel, or a specific Panel component ID. Unknown or inactive targets return PANEL_APP_NOT_FOUND before a UI display request is emitted.

app invoke <app-id> <action-name> --input '<json>' calls an Action on an enabled installed App, rather than a source package. It returns the call ID, trace ID, data version, and verification-record ID. Use app verification [--acceptance <id>] [--app <id>] [--limit <n>] to inspect the corresponding redacted, persisted runtime facts; add --json for machine-readable output.

app acceptance contract|status|export reads the single Portable Runtime acceptance registry used by the product, server, CLI, and release gate. status evaluates evidence against the active product version, runtime version, runner fingerprint, and contract fingerprint; only current-passed means the evidence is current. export always writes the complete machine-readable status document. Use --locale en for English presentation and --app <id> only when inspecting a non-default acceptance App.

For Apps that declare Secret slots, use app secrets inspect <app-id> to see required configuration without revealing values. Bind a declared slot with app secrets bind <app-id> --slot <slot> --source env|file|exec --id <secret-id> [--provider <provider>], then run app secrets verify <app-id>. A required unbound or unresolved slot leaves the App in needs-configuration and blocks enable with a SECRET_* error code. app secrets unbind removes the active Secret permission; retaining App data never retains Secret bindings.

Installing an App never grants its declared directory scopes automatically. Use app permissions inspect <app-id> to inspect them, app permissions document grant <app-id> --scope <scope-id> --path <directory> --mode read|read-write to grant or replace a runtime-host directory, and app permissions document revoke <app-id> --scope <scope-id> to revoke it. The effective mode cannot exceed the App's declaration, and every grant change immediately retires the previous runtime mount.

Automation guidance

  • When a query or management command supports --json, scripts and Agents should prefer the machine-readable output.
  • For non-interactive tasks, use nextclaw exec --format text|json|jsonl and interpret the documented exit codes.
  • Before changing configuration, secrets, host services, or permanently deleting data, inspect the command's --help output and permission boundary.
  • The CLI, UI, and built-in AI tools should call the same product owner. If the same capability behaves differently across entry points, please report it.

Resource protocol

CommandDescription
nextclaw resources listDiscover registered categories and objects from the running service; supports --type, --query and --limit; JSON output
nextclaw resources resolveResolve a real object URI into an immutable asset snapshot reference without executing its content

External collaboration

See external collaboration for setup and user journeys. Requires Node.js 22.13+.

CommandPurpose
nextclaw collaboration connectAdd a platform connection using existing local login
nextclaw collaboration presentationConfigure caller-owned reply presentation
nextclaw collaboration webhookConfigure GitHub webhook relay or restore polling
nextclaw collaboration install-adapterRegister an explicitly installed trusted adapter module
nextclaw collaboration adaptersList builtin and explicitly installed source adapters
nextclaw collaboration trustTrust a peer agent identity on a specific platform account
nextclaw collaboration migrate-discussionImport stopped official listener bindings and checkpoint without replay
nextclaw collaboration bindExplicitly repair a binding using an existing idle Codex task
nextclaw collaboration reconcileRecheck unknown executions without resubmitting work
nextclaw collaboration retry-runExplicitly retry failed/cancelled work after inspecting side effects
nextclaw collaboration statusShow connections, tasks and runtime without starting a model
nextclaw collaboration checkCheck configured source authentication and identity
nextclaw collaboration showInspect a task, its results and delivery failures
nextclaw collaboration ingestDurably accept a CloudEvent from a trusted local producer
nextclaw collaboration serve-eventsServe authenticated protocol events (optional; platform polling needs no server)
nextclaw collaboration resolve-outputReconcile uncertain delivery; retry or discard only after manual verification
nextclaw collaboration followExplicitly invite this agent to an existing source object
nextclaw collaboration runRun the local host in the foreground
nextclaw collaboration startStart the local host in the background
nextclaw collaboration stopStop this host; preserve bindings and recoverable work
nextclaw collaboration restartRestart the shared host after confirmed shutdown
nextclaw collaboration controlQueue status, pause, resume or cancel for a context key

Released under the MIT License.