Admin Guide

Understand Admin Panel access, operating areas, and the recommended platform setup flow.

The Admin Panel centralizes platform-level account, model, asset, marketplace, observability, and system settings. Only users with the admin or super_admin platform role can enter it; other roles are redirected to the regular Studio experience.

The Admin Panel affects the platform, not just the currently selected personal or organization workspace. Confirm the scope before changing roles, system models, pricing, or safety settings.

Administrator roles

Both admin and super_admin can open the Admin Panel and use most management features. A small number of high-impact operations require super_admin, including manually running or rebuilding platform usage rollups.

Do not infer authorization from a static role matrix. Each server API validates the action independently. A visible action that returns 403 requires a higher-privileged account.

Admin Panel areas

Group	Areas
Overview	Inbox for anomalies, pending reviews, quota, and configuration gaps
Users	Users, Waitlist, and Rewards
Organizations	Organization workspaces and member distribution
Models	Models, System models, and Marketplace
System	License, Playground, system time, websearch, and asset inventory
Data and operations	Insights, data platform, and deployment

Some legacy URLs remain compatible, but standalone SQL DB and On-prem functions are now folded into Data platform or Deploy.

Inbox

The Admin Panel opens on Inbox, which summarizes signals that need attention:

Pending users and marketplace submissions.
Usage, quota, or guardrail anomalies.
Missing system-model or service configuration.
Links to the corresponding management area.

Knowledge base ingest queue

Inbox shows a live ingest queue depth spanning document parsing, per-page parsing, and index finalization, with two numbers only: Queued (waiting) and In progress (running).

This line answers "what is the system doing right now," so it does not show a success count — a successful job simply makes the queue shorter. Failed ingest jobs live in the anomalies section, where you can re-process them directly. Re-processing a document fans it out into many per-page jobs, so a temporary spike in the queued number is expected.

Inbox is not a complete audit log. Continue an investigation in Users, Marketplace, Insights, or Data platform for full detail.

Models and system models

Models

Models maintains Model Zoo base models, including catalog metadata, public pricing, capabilities, upstream connections, routing priority, and enabled state. Changes can affect custom models that use the base model.

System models

System models supplies shared platform models and runtime defaults, including:

VLM and STT.
Runtime routing and basic-task models.
Guardrails and the evaluation judge.
Data Foundry and OpenClaw models.
Knowledge-base chunking, retrieval, and related platform tasks.

Each section reports whether its configuration is complete. A missing model affects only features that depend on it; behavior also depends on file type, feature switches, and available fallback paths.

Other core areas

Users: roles, account state, role quotas, per-user overrides, wallets, and test accounts.
Waitlist: OpenClaw onboarding and requested Model Zoo models.
Rewards: signup bonuses, referral rewards, and gift codes.
Organizations: organization workspaces and membership.
Marketplace: knowledge-base and skill review, listing fees, and platform commission.
Assets: platform inventory of knowledge bases, custom models, skills, and MCP tools.
Insights: requests, tokens, spend, performance, and request logs.
Data platform: request-log pipeline, ClickHouse, retention, cold archive, and data connections.
Deploy: deployment availability, machines and clusters, templates, and related settings.

Recommended initial setup

Verify the first super_admin account and create any additional administrators.

Under Users → Role presets, set the registration role, knowledge-base quotas, and group limits.

Create base models, upstream connections, and public pricing under Models, then verify Model Zoo.

Configure the VLM, STT, judge, Data Foundry, and knowledge-base task models that your installation uses.

Configure license, system time, websearch, data platform, and deployment features for the environment.

If billing is enabled, set Stripe environment variables, model pricing, and reward policy, then complete a test payment.

Use a test account to verify login, custom-model creation, API calls, billing, and permissions.

On this page