Commands Reference

Use slash commands at the > prompt. Every command also accepts the /jdai- prefix for compatibility (example: /jdai-help, /jdai-models).

Command groups

JD.AI currently supports 35+ interactive commands:

• Discovery and model/provider management
• Session and context controls
• Safety and execution controls
• Reviews and diagnostics
• Local models and MCP server management
• UX customization and persisted config
• Agent/hook profiles and project memory
• Skills lifecycle inspection and reload
• Workflows and checkpoints

Discovery and model/provider commands

/help

Shows all available commands and short descriptions.

/models

Lists available models and marks the active session model.

/model [current|list|set <id>]

Session-scoped model management:

• /model or /model current shows the active model for the current session.
• /model list lists available models.
• /model set <id-or-alias> switches only the current session model (does not persist defaults).

/model current
/model list
/model set claude-sonnet
/model set openai/gpt-4o

/model search <query>

Search remote model catalogs (Ollama, HuggingFace, Foundry Local) for models matching the query.

/model search llama 70b
/model search phi-3

Supports optional filters:

/model search --provider openrouter --cap reasoning --sort cost llama
/model search --sort popularity gpt

/model url <url>

Pull a model directly from a URL. Supports Ollama model URLs, HuggingFace repository URLs, and direct GGUF download links.

/model url https://ollama.com/library/llama3.2
/model url https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF

/providers

Lists detected providers and status.

/provider

Manage and inspect providers. Without arguments, shows an interactive picker to switch between available providers.

Subcommand	Description
/provider list	Show all providers with status, model count, and auth method
/provider add <name>	Interactive wizard to configure an API-key provider
/provider remove <name>	Remove stored credentials for a provider
/provider test [name]	Test provider connectivity (all or specific)

Provider names for /provider add: openai, azure-openai, anthropic, google-gemini, mistral, bedrock, huggingface, openai-compat.

Defaults management

/default

Show the current default provider and model (global and per-project).

/default provider <name>

Set the global default provider.

/default model <id>

Set the global default model.

/default project provider <name>

Set the per-project default provider for the current repo.

/default project model <id>

Set the per-project default model for the current repo.

/update [status|check|plan [target|latest]|apply [target|latest]]

Update orchestration commands:

• status — show update workflow config and enabled components
• check — check for available updates
• plan — print the ordered drain/update/restart/reconnect plan
• apply — execute the update workflow (subject to approval policy when enabled)

/doctor

Runs environment diagnostics (runtime, providers, model count, tools, plugin load status, and CLI availability).

/docs [topic]

Shows links to JD.AI documentation. Without a topic, lists key doc areas.

Session and context commands

/clear

Clears current conversation history for the active session.

/compact

Triggers context compaction immediately.

/reasoning [auto|none|low|medium|high|max]

Shows or sets session reasoning effort override for compatible reasoning models.

/reasoning
/reasoning low
/reasoning max
/reasoning auto

/context

Shows current context window usage with a visual bar.

/cost

Shows current session token usage.

/sessions

Lists recent sessions for the current project hash.

/resume [id]

Resumes a previous session. With no ID, shows a list and usage hint.

/name <name>

Sets a friendly name on the current session.

/history

Shows a summarized turn history for the current session.

/export

Exports current session to JSON under the JD.AI data directory.

/fork [name]

Forks the active session into a new session branch, optionally named.

Safety and execution controls

/autorun [on|off]

Enables or disables auto-run behavior for tools.

Warning

on can allow writes, shell commands, and networked operations to execute without per-call confirmation.

/permissions [on|off]

Enables or disables permission checks entirely for the current session.

Warning

off is equivalent to running with fully skipped confirmation behavior for that session.

/sandbox

Shows available sandbox modes and configuration guidance.

/plan

Toggles plan mode on/off for exploration-first behavior.

/copy

Copies the last assistant response to clipboard.

/diff

Shows uncommitted git diff (git diff, then staged diff fallback).

Audit commands

/audit

Shows recent audit events from the in-memory audit buffer. Supports filtering by severity. Use this command to inspect what actions the agent has taken, what tools were invoked, and whether any policy denials occurred.

/audit
/audit --severity warning
/audit --limit 50

Parameter	Description
--severity	Minimum severity level to display (debug, info, warning, error, critical)
--limit	Number of events to show (default: 20, max: 100)

Output format:

Audit Log (142 total events, showing 20)

.. 14:22:07 tool.invoke [read_file] — Debug
.. 14:22:08 tool.invoke [grep] — Debug
~~ 14:22:09 tool.invoke [run_command] — Warning
-- 14:22:10 session.create — Info

Severity icons in the output:

Icon	Severity
..	Debug
--	Info
~~	Warning
**	Error
!!	Critical

For interactive filtering and programmatic access, see the Gateway API Reference and Audit Logging.

Review and security commands

/review [--branch <name> --target <name>]

Runs model-assisted code review over local changes or a branch comparison.

/review
/review --branch feature/login --target main

/security-review [--full] [--branch <name> --target <name>]

Runs OWASP/CWE-oriented heuristic security scanning.

• Default mode scans changed files.
• --full scans tracked repository files.

/security-review
/security-review --full
/security-review --branch hotfix/auth --target main

Local model commands

/local list

Lists registered local GGUF models.

/local add <path>

Registers a .gguf file or scans a directory.

/local scan [dir]

Scans directory for local models (defaults to JD.AI model dir).

/local remove <model-id>

Removes a model from the local registry (does not delete model file).

/local search <query>

Searches HuggingFace for GGUF repositories.

/local download <repo-id>

Downloads GGUF model from HuggingFace and registers it.

See Local Models for setup and storage behavior.

MCP server commands

/mcp list

Lists configured MCP servers and current status.

/mcp add <name> --transport stdio --command <cmd> [--args ...]

Adds a stdio MCP server.

/mcp add <name> --transport http <url>

Adds an HTTP MCP server.

/mcp remove <name>

Removes a JD.AI-managed MCP server entry.

/mcp enable <name>

Enables a configured MCP server.

/mcp disable <name>

Disables a configured MCP server.

Workflow and checkpoint commands

/workflow [list|show|export|replay|refine]

Manages workflow artifacts captured from multi-step execution.

/workflow list
/workflow show onboarding
/workflow export onboarding mermaid
/workflow replay onboarding
/workflow refine onboarding

/checkpoint [list|create|restore <id>|clear]

Manages safety checkpoints for rollback.

/checkpoint list
/checkpoint create before-refactor
/checkpoint restore abc123
/checkpoint clear

UX, settings, and stats commands

/spinner [none|minimal|normal|rich|nerdy]

Sets turn progress spinner style.

/theme [name]

Lists or sets terminal theme.

Supported aliases: default, default-dark, monokai, solarized-dark, solarized-light, nord, dracula, one-dark, catppuccin, catppuccin-mocha, gruvbox, high-contrast.

/vim [on|off]

Toggles vim editing mode in the interactive prompt.

/output-style [rich|plain|compact|json]

Sets renderer output mode. json is session-only and is not persisted as the startup default.

/stats [--history|--daily]

Shows session usage stats, cross-session stats, or daily aggregates.

/config [list|get <key>|set <key> <value>]

Reads or writes persisted command-facing settings.

Supported keys: theme, vim_mode, output_style, spinner_style, prompt_cache, prompt_cache_ttl, autorun, permissions, plan_mode, welcome_model_summary, welcome_services, welcome_cwd, welcome_version, welcome_motd, welcome_motd_url.

For cache behavior details and tuning guidance, see Prompt Caching.

/init

Creates a starter JDAI.md in the current project if missing.

/instructions

Displays merged instruction sources loaded into the prompt.

/plugins

Manages installed Plugin SDK packages and runtime lifecycle.

Subcommands:

• /plugins list — list installed plugins with enabled/loaded status
• /plugins install <path-or-url> — install and enable plugin package
• /plugins enable <id> — enable plugin
• /plugins disable <id> — disable plugin
• /plugins update [id] — update one plugin or all installed plugins
• /plugins uninstall <id> — uninstall plugin package
• /plugins info <id> — show plugin details and last error

/plugins
/plugins install https://example.com/My.Plugin.1.2.0.nupkg
/plugins update my-plugin
/plugins uninstall my-plugin

/skills [status|reload]

Shows managed skill lifecycle state and supports on-demand refresh.

• /skills or /skills status prints deterministic skill eligibility status (active, excluded, shadowed, invalid) with reason codes.
• /skills reload forces a refresh from source directories and skills.json config.

/skills
/skills status
/skills reload

Profile and memory commands

/agents [list|create|delete|set]

Manages local agent profiles in ~/.jdai/agents.json.

/agents list
/agents create reviewer
/agents set reviewer model gpt-4o
/agents set reviewer tools read_file,grep,git_diff

/hooks [list|create|delete|toggle|set]

Manages local hook profiles in ~/.jdai/hooks.json.

/hooks list
/hooks create lint-on-save
/hooks set lint-on-save command dotnet format --verify-no-changes
/hooks toggle lint-on-save

/memory [show|edit|append|reset]

Views or updates project memory file (JDAI.md) in the current repository.

/memory show
/memory append Prefer xUnit for unit tests
/memory reset

Exit commands

/quit

/exit

Ends the JD.AI interactive session.

Gateway channel commands

Gateway adapters expose command variants natively per platform:

• Discord: slash commands (example: /jdai-help)
• Slack: slash commands (example: /jdai-help)
• Signal: command prefix form (example: !jdai-help)
• OpenClaw bridge: message commands (example: /jdai-help)

jdai-help

Lists gateway command usage.

jdai-usage

Shows uptime, turn counts, and per-agent usage.

jdai-status

Shows current agent/channel status.

jdai-models

Lists providers/models available to the gateway runtime.

jdai-switch <model> [provider]

Switches spawned agent model/provider.

jdai-clear [agent]

Clears one agent conversation or all if omitted.

jdai-agents

Lists active gateway agents and route mappings.

jdai-route [agent]

Shows the current channel route or remaps the current channel to a matching agent/provider/model.

jdai-route
jdai-route ollama
jdai-route gpt-5.3-codex

jdai-routes

Lists all channel-to-agent mappings.

jdai-provider [name]

Shows current mapped agent provider or switches the mapped channel to a newly spawned agent on the selected provider.

jdai-provider
jdai-provider openai
jdai-provider ollama

jdai-providers

Lists detected provider availability and models from the gateway runtime.

OpenClaw bridge invocation

When issuing gateway commands from an OpenClaw conversation, always use /jdai- prefix messages:

/jdai-routes
/jdai-route ollama
/jdai-provider openai

Quick reference

Command	Purpose
/help	Command list
/models, /model, /model search, /model url	Model browsing, switching, search, and pull by URL
/providers, /provider, /default ...	Provider discovery, credential management, and defaults
/sessions, /resume, /name, /fork	Session lifecycle
/clear, /compact, /context, /cost, /history, /export	Context/session operations
/autorun, /permissions, /sandbox, /plan	Safety and execution mode
/audit	Inspect audit events with optional severity filter
/review, /security-review	Code and security review
/local ...	Local model operations
/mcp ...	MCP server management
/checkpoint ..., /workflow ...	Checkpointing and workflows
/theme, /vim, /spinner, /output-style, /stats, /config	UX and runtime settings
/agents, /hooks, /memory, /init, /instructions, /plugins, /skills, /doctor, /docs	Project and profile operations
/quit, /exit	Exit session