Config Center

Config Center is Vibex’s configuration entry point, opened from the Config Center button in the sidebar or from the system tray. It brings Agent runtimes, model Providers, MCP servers, Skills, and reusable Prompts together and shows status and validation results for each resource.

​

Sections

The Config Center currently has five primary sections:

​

Common operations

​

Search and filter

Each section has a search field at the top of its left list. Agents match by name, id, and description; Providers by name, address, model, and key; MCP, Skills, and Prompts by name or description. The list remembers your ordering and filters until you change them.

​

Save and validate

1. Fill in the fields and save.
2. Click Validate on the resource card. MCP and Skill validation is offline: it checks required fields, format, and scope, but starts no processes and makes no network requests.
3. For a model Provider you can also run Test connection, which makes one real call to the model API.

​

Scope

Some resources let you set a scope:

• User: visible to all your projects.
• Workspace: visible only to the current workspace.
• Project: follows project conventions where supported.
• Global: a global default; use sparingly.

Start with the smallest scope. MCP servers currently offer only User and Workspace.

​

Per-Agent enablement

Resources (MCP, Skills, Prompts, Providers) relate to Agents through per-Agent enablement. The Agent enablement area below a resource card provides a switch for each added Agent; only Agents with the switch on load the resource. When you no longer need a resource, disable it first instead of deleting it right away—disabling keeps the record so you can restore it later. Deleting a resource removes the configuration but not files already written into an Agent’s native directory; use the export preview’s rollback for those.

​

Credentials

Vibex stores credentials through secret references or the operating system’s secure storage. The desktop prefers the system keychain; a host without a usable keychain falls back to a file only its owner can read.

• Saved keys are masked by default; the eye icon reveals a value temporarily.
• A field you did not change does not rewrite the saved value; saving it blank keeps the old value.
• To remove a credential, use the explicit Clear action, not an empty field.
• Never paste tokens, private keys, full environment variables, prompts, or file contents into descriptions or diagnostic logs.

​

Agents

The list is grouped by state: enabled, added but disabled, and not yet added. An Agent’s details contain:

• Agent installation: install, check for updates, upgrade, roll back to the pinned version, or uninstall the managed runtime.
• Native credentials: sign in to the Agent’s own account, refresh auth methods, verify the account, refresh models, re-authenticate, or sign out.
• Model Provider profiles: this Agent’s own addresses, credentials, models, and reasoning depth.
• Built-in tools: per-Agent switches for the terminal, browser, computer use, and other tool panels.

See Agents and provider profiles.

​

Providers

The Providers page lists every shared Provider once and shows which Agents use it. The pinned row at the top is the Gateway; the sidebar has Providers and Route groups tabs.

• Providers: address, credentials, keys, models, request headers, proxy, and request body fields.
• Route groups: compose several Providers or models into one Agent-facing name, with strategy, failover, limits, and visibility.
• Gateway: the loopback endpoint that routes, translates protocols, cools down keys, retries, and records the request ledger.

See Providers, route groups, and the local gateway.

​

MCP and Skills

Both sections open on their market by default:

• The MCP market shows the transport, command, environment variables, and request headers; review the command before installing.
• The Skill market shows the SKILL.md body, file list, install count, and source; read the full document and choose which Agents receive it.
• Installed entries offer Reinstall; removed entries explain themselves first.
• You can also create a resource or import an existing MCP server or Skill.

See MCP and Skills.

​

Prompts

Prompts stores reusable prompt bodies. After saving, type / in the composer to search and insert one; a Prompt can also be enabled per Agent. A Prompt is a piece of text. A Skill is a capability package with files and instructions. Use a Skill when the Agent needs tools or domain knowledge; use a Prompt when you keep typing the same instruction.

​

Export to Agent-native configuration

MCP, Skills, and Prompts can be written into an Agent’s own configuration files so external CLIs see the same setup. On a resource card, use Preview export → Apply → Rollback:

1. Preview export lists the files that will change or be created.
2. Review them, then Apply.
3. If something goes wrong, Rollback from the same place.

The target depends on the Agent, for example Codex’s ~/.codex/config.toml, Claude’s ~/.claude/settings.json, and each Agent’s Skills directory.

​

When configuration goes wrong

1. Confirm the resource’s scope and enablement state.
2. Run Validate to see whether there are structured errors.
3. Confirm the resource is switched on in the target Agent’s Agent enablement.
4. Check which authoritative runtime you are connected to—local and remote runtimes keep separate configuration.
5. If state is inconsistent after a restart, see Troubleshooting. Do not delete the SQLite database directly.