Commands

Synopsis

Run central with no command to open the interactive interface. To see the flags of a single command, run central <command> --help.

This reference uses the following notation:

<value>

A value you provide.

[value]

Optional — can be omitted.

value...

Can be repeated.

Agent commands

central run

Run an agent with JetBrains Central CLI proxy credentials injected as environment variables. The proxy starts automatically if it is not already running. Anything after -- is passed straight to the agent:

central run <agent> [-- extra-args...]

Example:

$ central run codex -- "explain this code"

For copilot, --provider is required on every run, and --model is optional:

central run copilot --provider=<openai|anthropic> [--model=<id>] [-- extra-args...]

Example:

$ central run copilot --provider=anthropic -- -p "explain this diff"

central add

Configure an agent to route requests through the JetBrains Central CLI proxy. This modifies the agent's configuration file permanently. If that file already exists, a timestamped copy is saved next to it before changes are applied.

$ central add <agent>

Example:

$ central add claude

Supported agent names are claude, codex, gemini, junie (alias junie-cli), pi (alias pi-dev), and copilot.

For copilot, --provider is required and --model is optional. Without --model, the default is gpt-5.6-luna for OpenAI and claude-opus-5 for Anthropic:

$ central add copilot --provider=<openai|anthropic> [--model=<id>]

Instead of editing a configuration file, central add copilot sets Copilot's environment variables, so no backup is saved. Open a new terminal afterwards.

central remove

Remove JetBrains Central CLI configuration from an agent, so that it reaches its provider directly again.

$ central remove <agent>

Example:

$ central remove claude

central repair

Recover your JetBrains Central CLI setup in one step, for example after you change the proxy port or when agents stop working. The command takes no arguments:

$ central repair

Repair:

• Refreshes all the credentials JetBrains Central CLI has stored, and asks you to log in again if they can no longer be refreshed.

• Fetches your organization's settings again, including which agents you can use.

• Removes deprecated shell aliases and connects the agents they pointed to.

• Removes the legacy ~/.wire data directory, if it is still there.

• Reconnects every connected agent. GitHub Copilot keeps the provider and model you selected.

• Restarts the proxy, so that the changes take effect.

The output lists the steps that were taken. Refreshing credentials and settings, reconnecting agents, and restarting the proxy always appear; optional steps that aren't needed, such as removing the legacy data directory, are left out.

Each step is best-effort: if one fails, repair reports it and continues with the rest. You can run repair as often as you need. If JetBrains Central CLI can't read GitHub Copilot's provider and model, repair disconnects Copilot; connect it again with central add copilot.

Repair is not available in the interactive interface.

central agents

List all supported agents and their current status.

$ central agents

Every agent is listed under a single Agents heading. Agents blocked by AI governance are dimmed and their status reads "blocked by AI governance". Agents the current session cannot use are dimmed with the reason instead — "not supported with service accounts", or "not supported with the bedrock protocol" — since a login method or a missing capability is not a policy decision.

To refresh provisioned AI agent settings manually:

$ central agents refresh

This first checks organization feature flags from JetBrains Cloud using org_user_token as a bearer token. It downloads provisioning settings only when both ai-seats and ai-paid-features are enabled; otherwise it stores unrestricted settings (settings: null). The provisioning settings include the agent allow-list and the MCP servers that your organization provides.

MCP policy commands

These commands control which of the MCP servers that your organization provides are added to your agents. allow and deny accept the full server ID or the exact server name, as listed by central policies. Names are case-sensitive. If several servers have the same name, use the ID.

central policies

List your organization's MCP servers with their ID, status (allowed, pending, or denied), and connection details.

$ central policies

central policies allow

Allow an MCP server, so JetBrains Central CLI adds it to your agents. With -a, allow all pending servers at once; denied servers stay denied. -a cannot be combined with an ID or name.

central policies allow <id-or-name> central policies allow -a

Examples:

$ central policies allow 0f6c2e8a-1b3d-4c5e-9f7a-2d8b6e4c1a90 $ central policies allow "Issue tracker"

central policies deny

Deny an MCP server, so JetBrains Central CLI removes it from your agents. The server stays denied until you allow it again.

$ central policies deny <id-or-name>

Authentication commands

central login

Authenticate with JetBrains Account. Opens a browser for OAuth 2.0 login.

$ central login

Alternatively, you can log in to a JetBrains Central Server instance and use AI Enterprise. If a Central Server login is triggered while you are already logged in with your JetBrains Account, a new login is started.

$ central --enterprise # Log in and open the menu $ central login --enterprise # Log in

Flag	Description	Default
--enterprise (alias: --server)	Authenticate against a JetBrains Central Server instance instead of JetBrains Account. Prompts interactively for the server URL.	false

central logout

Clear credentials, delete the cached AI governance policy, and stop the proxy. The configuration file is preserved, so agents stay connected.

$ central logout

Status and info commands

central status

Show authentication, proxy, and agent status.

$ central status ⣿ Auth logged in (AI Pro) ⣿ Proxy running on :19516 ⣿ Agents Claude Agent, Codex

central limit

Show your current AI usage limit for seat-based access, or quota for license-based access, reported in AI credits. central quota is kept as an alias.

$ central limit

central access

Show enabled AI access sources for the logged-in user. Workspaces are listed first, followed by licenses. Passing a source ID switches the active AI access source. Workspace IDs are not shown in the list; license IDs are shown in the details after the licensed-to name.

$ central access $ central access workspace:org-123:ws-456 $ central access license:L5DSO53F1QP

central licenses

A hidden temporary compatibility alias for central access. It prints a warning that license-based management was superseded by AI access-based management, then runs the access command.

$ central licenses

central log

View the local request log.

$ central log

central --version and central --help

$ central --version central 0.1.20 (commit: 7911b2e, built: 2025-02-08T10:30:00Z)

central --help (or -h) shows usage information. -v is a short form of --version.

Proxy commands

Read more about the proxy.

central proxy start

Start the proxy in the background and exit.

central proxy start Proxy started on port 19516 (PID 12345)

The proxy runs as a daemon process. It survives terminal close and continues until explicitly stopped.

central proxy stop

Stop the running proxy and exit.

$ central proxy stop ✓ Proxy stopped

Maintenance commands

central config

View the current JetBrains Central CLI configuration.

$ central config

To change a value, use central config set <key> <value>:

$ central config set log-level debug $ central config set port 19517 # follow a release channel, or pin a specific version $ central config set release-track stable $ central config set release-track 1.2.3 # alias for license-backed access $ central config set license-id A1B2C3D4E5 $ central config set google-analytics on $ central config set google-analytics off $ central config set jumpscare-mode on $ central config set jumpscare-mode off # approve MCP servers automatically, or review them first $ central config set approve-all-policies on $ central config set approve-all-policies off

google-analytics has no default: it stays not set until you make a choice, and nothing is sent while it is not set. jumpscare-mode is on by default, so starting the proxy while you are logged out opens the browser login page. approve-all-policies is on by default, so MCP servers are approved automatically.

central update

Resolve the update target. The precedence is --target-version, then the configured release-track — either a channel's latest published version or a pinned version.

$ central update # skip the prompt $ central update --install # install a specific version $ central update --target-version 1.2.3 --install

central uninstall

Remove JetBrains Central CLI and all its data. You are asked to confirm unless --yes is passed.

$ central uninstall $ central uninstall --yes # skip the prompt

Environment variables

The environment variable takes precedence over the port saved in ~/.jetbrains-central/config.json.

Exit codes

Interactive interface

The default entry point. It guides you through login, agent setup, and proxy management. On first run it walks you through the login process; on subsequent runs it opens the main menu directly.