AFK is a browser-based command center for engineering teams running coding agents. It gives teams dashboards, persistent sessions, review gates, automations, MCP tools, and bring-your-own-model routing.

AFK is built for engineering teams, platform teams, and engineering leaders who need visibility, control, and continuity around coding-agent work rather than isolated terminal sessions.

Does AFK replace model providers?

No. AFK is model-agnostic and bring-your-own-key. Teams keep using their preferred providers, private endpoints, subscriptions, or local OpenAI-compatible models without AFK token markup.

For engineering teams · agent operations

The command center for coding agents.

Run more agent work in parallel, let it continue safely while you're away, and return to one browser view when a decision, approval, or intervention needs you. AFK keeps every session visible, steerable, and under your control.

Start free — 60 days full access See the system See pricing

Every tier, every feature, 60 days free. No token markup.

Live control plane · Session memory · Agent teams · Webhook triggers · Review gates · BYOK routing · No token markup

Live control

Zoom from team-wide signal into the exact agent run.

The dashboard shows where attention is needed across parallel work. Open the exact session to inspect plans and tool calls, steer its direction, approve a gated action, or stop it. Work can continue while you're away without becoming invisible or uncontrolled.

Approval gates and plan mode for human-in-the-loop control
Tool trail, context, and cost visibility in every session
Persistent, resumable agent sessions with durable operational memory

How it works

Launch. Remember. Coordinate. Resume.

Trigger

Start sessions from the browser, connected GitHub repositories, or inbound webhooks from CI, monitoring, ticketing, and internal systems.

Remember

Prompts, plans, tool calls, outputs, costs, context, and handoffs are preserved as durable operational memory for the next turn.

Coordinate

Spawn specialist sub-agents, delegate focused work, exchange messages, track background tasks, and bring results back into the parent session.

Resume

Long-running work survives reconnects and handoffs. AFK preserves the full session memory — not just the last prompt.

Proof that shows up in real work

Not a leaderboard. Four outcomes teams can feel.

AFK does not market model IQ. It markets the operating surface around agent work — always-on execution, shared visibility, reviewable mutations, and sessions that remain useful after you leave.

Close the laptop

Work keeps running when you leave. The daemon keeps the agent alive so the job does not die with a tab or a machine sleep.

Second human joins

Another teammate can open the same live run from the browser dashboard and inspect status, tool trail, and decisions.

Gate before land

Permission modes, plan review, and tool gates pause risky mutations until a human confirms or steers.

Resume with memory

Prompts, plans, tools, costs, and handoffs remain as operational memory so the next turn continues from what actually happened.

AFK complements tools like Claude Code and Cursor. Those focus on what an agent can do. AFK focuses on what a team can see, control, reuse, and inspect while agents work.

No token markup

Bring your own keys. Keep the costs visible.

AFK is not a model provider. Use hosted APIs, private endpoints, or self-hosted OpenAI-compatible models. AFK charges for the operating surface — visibility, coordination, durable sessions, and the agent harness.

Pay providers at their rates
Pay AFK for orchestration
Switch providers without switching workflows

Start free — 60 days See pricing