Kolpolok AI Mission Control

Monitor LLM usage, spend, performance, and activity across your organization.

Uses your own linked LiteLLM key automatically — nothing to paste. Conversations are saved automatically to your account.

Engineering knowledge, solved problems, references, and operational intelligence.

AI-compiled, human-reviewed capabilities built from your Knowledge Base — versioned and exportable in a format your OpenClaw agents can use directly.

Company/team context — some you write directly, some Palette keeps refreshed from your Knowledge Base, some just linked out to where it already lives (Notion, Drive, etc).

Hand off a decision, its next steps, and whatever context comes with it — to a person, a team, or an agent — instead of starting over.

Merged across every API key they've ever held.

Each team is an access profile — its own models, budget, and department.

Manage user accounts, project access, model permissions, and LiteLLM credentials.

Manage user and project wallets, balances, funding, and wallet activity. Real credit ledger, immutable transaction history.

A team-agnostic tag for "which piece of work this usage was for" — attach one to a chat call (see the Chat tab) to attribute its cost. Anyone can select a project; creating/archiving one needs a team-lead or superadmin.

Your own personal tool credentials — connect one here, and any agent acting on your behalf can use it without you reconnecting it separately each time. This is the one view where you only ever manage your own data, not your team's.

Every real admin mutation — who granted a key, when, and why. Superadmin-only.

Turn model connections on or off org-wide — a disabled model disappears from the Playground, Projects, and Users model pickers, and is refused if called directly. Nothing here deletes a model, its provider key, or any historical usage. Superadmin-only.

Which source IPs may send chat requests (API keys and the Playground). Rules apply to everyone, or to one user, project, agent or API key. A block always wins; otherwise the most specific target with an allowlist decides (key → agent → project → user → global). The dashboard itself is never IP-restricted. Superadmin-only.

Non-human identities — a bot/service integration that makes LLM requests through this same gateway, owned by one user or one project. Never a dashboard login of its own. Superadmin-only to create/manage; owners and project members can view their own.

Every real key ever generated — who holds it, which project it's scoped to, and what it's actually spent. Search here first if you suspect a key is compromised. Superadmin-only.

Your own API key usage only — real settled cost, never estimates.