> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agentbees.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Why AgentBees

> An enterprise-safe workspace for running many coding agents in parallel — on your cloud, under your policy, billed by what you actually use.

**AgentBees is a coordination layer over the coding-agent CLIs you already
trust** — Claude Code, Codex, Gemini, Grok, Kiro — running many of them **in
parallel** on your own repos, in isolated cloud workspaces. Not another IDE.
Not another agent. The place your team runs those agents at scale, with real
gates and a real audit trail.

<Note>
  If you're comparing us to "AI in your IDE" tools: those run **one** agent at a time on **your
  laptop**. AgentBees runs **N**, in parallel, on **your cloud**, with team-wide review + audit +
  billing.
</Note>

## The five differentiators

<CardGroup cols={2}>
  <Card title="Parallel, not serial" icon="grid-2">
    Fan out one task to N agents/models in isolated workspaces, compare diffs, ship the winner. This
    is the wedge — every other "AI IDE" runs one agent at a time.
  </Card>

  <Card title="Any CLI agent, one control plane" icon="terminal">
    Claude, Codex, Gemini, Grok, Kiro — same task surface, same diff/PR flow, same billing, same
    audit trail.
  </Card>

  <Card title="Fast, disposable workspaces" icon="bolt">
    Sub-second pod boot, one container per agent, per-task checkpoints + point-in-time restore. A
    task, not a laptop, is the unit of work.
  </Card>

  <Card title="Enterprise-safe by construction" icon="shield-halved">
    Internal AI gateway (org creds, no per-user keys), OrgPolicy + zero-retention, SSO-brokered
    credentials, tenant isolation, per-tenant cost attribution & chargeback.
  </Card>
</CardGroup>

<Card title="HoneyBox skills + shareable runs" icon="share-nodes">
  Reusable, org-scoped agent playbooks (auto-inject or on-demand) plus opt-in `/share` read-only
  links, so wins **compound across the team** instead of living in one person's shell.
</Card>

## Beyond a single task: multi-agent pipelines

A single task is just the start. Compose a **pipeline** of bees —
`jira → design → dev → QE → PR → security` — with human gates between stages,
a **verdict-gated remediation loop** (max 3 rounds) when QE or security fails,
and per-lane fan-out for AIDLC-style breakdowns. This is where most competitors
stop at "one agent, one task."

See [Pipelines](/workflows/pipelines) for how to compose one and
[Flow](/workflows/flow) for the mission-control view of every run in your org.

## Runs where your code lives

* **AWS EKS today** — shared pool namespace, dedicated per-tenant namespace,
  or dedicated cluster SKUs.
* **GCP + Azure on the roadmap.**
* **BYO-compute runner on the backlog** for orgs that want agents inside their
  own VPC — the platform ships the runner, they host it.

## The pitch grid

| Question a serious buyer asks                  | The one-line answer                                                                                                              |
| ---------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------- |
| What is it?                                    | A parallel-agent workspace for coding — you keep the CLI agents you trust, we run them on your cloud in isolation.               |
| Why not just use Claude Code / Cursor locally? | Because one laptop runs one agent. We run N, in parallel, side-by-side, with team-wide review + audit.                           |
| Who owns the model keys?                       | You do. BYO provider keys, or route via your own gateway. Nothing sensitive touches our infra.                                   |
| What is a "workspace"?                         | An isolated Kubernetes pod (optional gVisor / Kata microVM), one git worktree, egress carveout to your repo + your gateway only. |
| What lives while a run happens?                | Only a 1-hour scoped GitHub App installation token, in env, never on disk. Dies with the pod.                                    |
| Billing?                                       | Flat monthly (Pro/Team), pay-per-use prepaid wallet, or purchased add-ons. All three coexist per-org.                            |
| Where does it run?                             | Your K8s cluster (AWS EKS today; GCP/Azure roadmap; BYO-runner backlog).                                                         |

## Next

<CardGroup cols={2}>
  <Card title="FAQ" icon="circle-question" href="/getting-started/faq">
    The isolation / tenancy / credentials / pricing questions serious buyers ask.
  </Card>

  <Card title="Onboarding" icon="rocket" href="/getting-started/onboarding">
    Sign in, connect a repo, run your first agent.
  </Card>
</CardGroup>