The mission command for agent fleets.
Contracted, gated, fully audited.
The Kubernetes
for AI Agents
$ man a8s · kubectl-for-agents · docs.agentnetes.io ↗
// FOR_WHOM
Three audiences. One platform.
For platform teams
Run an internal agent fleet for your own engineering organization. Self-host the binary, full sovereignty over data, Kubernetes-grade primitives you already understand.
- Single-binary install (~5 min)
- Embedded etcd; no Redis/Postgres
- Declarative state, self-healing controllers
For agencies
Serve 5–50 clients on one orchestrator with hard isolation. Workspace-per-client model, white-label dashboards, separate billing per workspace.
- Org → Workspace → Project hierarchy
- White-label dashboard with custom domain
- Audit log + usage metering per workspace
- Publish client-specialized agents (revenue share)
For non-engineers
Run CFO, Legal, COO, CHRO, CMO, CRO agents on your business processes. Brain-dump UI, plain-English playbooks, no infrastructure knowledge required.
- Describe a goal, agents decompose it
- Approval gates on irreversible actions
- Idea → production in hours
// MISSION_CAPABILITIES
System core modules
The three primitives every Agentnetes deployment runs: declarative state, real-time telemetry, and resource gating.
Declarative orchestration
Define your agent's desired state in YAML. Agentnetes handles provisioning, lifecycle management, and auto-healing without manual intervention. The same declarative loop Kubernetes runs for pods, applied to AI agents.
Real-time SSE telemetry
Low-latency streaming telemetry for every agent thought-process. Visualise decision trees and neural activations in high-fidelity dashboards.
Multi-tier resource gating
Granular control over LLM token usage, memory overhead, and external API budgets. Prevent runaway processes with hardware-level enforcement.
// FLEET_PREVIEW
Mission Control Dashboard
Monitor cluster health, agent throughput, and autonomous conflict resolution from a single glass-pane interface. Designed for high-density operations.
- Global Agent Latency Heatmaps
- Automated Token Burn-Rate Analysis
- Neural Trajectory Visualization
[08:22:02] AGENT_01: VECTOR_DB MATCH FOUND (ID: 9822)
[08:22:02] GATEWAY: TOKEN BUDGET APPROVED (COST: 0.00021)
[08:22:03] ORCHESTRATOR: REPLICATING INSTANCE TO EU-WEST-1
[08:22:04] SYSTEM: CPU LOAD BALANCED...
[08:22:05] AGENT_02: LISTENING ON PORT 8080...
// EXTENDED_CAPABILITIES
Operator Toolkit
Agent Fleet Management
Register agents with roles, capabilities, and model preferences. Monitor load, utilization, and costs per agent in real-time.
Mission Orchestration
Describe a goal in natural language. The orchestrator decomposes it into phased missions with task dependencies, then agents execute autonomously.
Execution Traces
Waterfall view of every agent step: planning, research, implementation. See exactly where tokens and dollars are spent.
Smart Scheduling
Filter, Score, Bind pipeline matches tasks to the best agent based on capabilities, load, and affinity. Like kube-scheduler for AI.
Human-in-the-Loop
Agents escalate to your inbox when stuck. Choose from pre-defined actions: retry, reassign, abort, or provide guidance.
Stall & Loop Detection
Monitor detects tasks stuck for >10 minutes or agents repeating the same output. Automatically escalates before costs spiral.
// METHODOLOGY, NOT MAGIC
We don’t start with AI.
We start with your outcome.
One orchestrated engine takes any business outcome — a CEO’s vision, a factory floor, a regulated trading desk — and drives it to a shipped solution. Software, hardware, third-party services, cloud, regulated data — whatever the problem actually needs. Agents do the work end-to-end across your existing stack — GitHub, Stripe, AWS, Kafka, Datadog, hardware fleets, the regulators. Humans unlock the gates that only humans can sign.
Anchor
the outcome question
// SYSTEM_ARCHITECTURE
Kubernetes-Grade Orchestration
You declare "I want this done" and controllers converge reality toward that state -- retrying, reassigning, and escalating automatically.
Agent goes offline? Task auto-reassigns. Execution stalls? Monitor detects and escalates. Output loops? Different agent retries.
etcd embedded, UI embedded, everything in one container. No Redis, no Postgres, no message queue. Just docker run.
Each agent can use a different LLM: Claude for reasoning, GPT-4 for code, Gemini for research, Ollama for local. Per-namespace credential scoping.
// PROTOCOL_INTEGRATIONS
Works with your stack.
Pulls tasks from your existing tools, dispatches them through your existing model providers, talks to other agents via open protocols. No re-platforming.
// MARKETPLACE
A vetted fleet you can hire in seconds.
Pro and Enterprise customers publish specialized agents to the Agentnetes marketplace — domain-deep agents we couldn’t have built ourselves. Every listing is reviewed by us before it ships. Hire on demand; pay the publisher; run it inside your namespace under your policies, your vault, your audit log.
Every published agent is reviewed against five pillars — capability, security, guard rails, observability, compliance — before it lists. No silent installs.
KYC + AML pipeline for regulated finance. Identity verification, sanctions screening, EDD escalations — outputs a packet your compliance officer can sign.
Predictive maintenance over OPC-UA + MQTT. Tracks drift, schedules interventions, opens work orders against your CMMS.
Static + symbolic analysis for Solidity. Flags reentrancy, integer overflow, and access-control gaps before mainnet.
Routes incoming claims by payer, plan, and clinical code; flags denials before submission. HIPAA-ready data handling.
Doc review at discovery scale — privilege flags, redaction proposals, exhibit indexing. Outputs a privilege log your firm can defend.
Parametric SolidWorks-style drawings from natural-language constraints; export to STEP / IGES / DXF for your CAM pipeline.
Pull a listed agent into your namespace; it runs under your policies, your vault, your audit log — never the publisher’s.
Any Pro or Enterprise user can package and publish. We review against the five pillars, then list. Set your own price; revenue is shared.
Private marketplaces — white-label, internal-only listings scoped to your org. Your agents stay yours, your IP stays yours.
// PRICING
Plans & Tiers
Same Kubernetes-grade core across every tier — the difference is scale, integrations, and assurance. We don’t quota LLM tokens; bring your own provider keys and pay them directly.
Engage us via the Orchestrator for deployment, customisation, and managed services on top of any plan.
Community
Solo devs, OSS, and learning. Run the platform locally with no per-developer cost.
- 3 concurrent agents · 3 namespaces
- Bring your own LLM keys
- GitHub + Linear integrations
- 90 days of audit log retention
- Community support
Pro
billed annually ($374.40/user/year) · save 20%
Indie teams & small consultancies running production fleets.
- Unlimited agents & namespaces
- Custom agents · self-hosted agents
- Publish agents to the marketplace
- Cost analytics (Autopilot dashboard)
- 1 year of audit log retention
- Email support
Enterprise
SSO, audit, and contractual SLAs for regulated orgs and large teams.
- SSO (SAML / OIDC)
- Multi-seat & org management
- Private marketplace (white-label, internal-only)
- Up to 7 years of audit retention (default 2y)
- Dedicated support & SLA
- Air-gapped / sovereign-cloud install
| Feature | Community | Pro | Enterprise |
|---|---|---|---|
| Concurrent agents | 3 | Unlimited | Unlimited |
| Namespaces | 3 | Unlimited | Unlimited |
| Bring your own LLM keys (Anthropic / OpenAI / etc.) | |||
| GitHub integration | |||
| Linear integration | |||
| Custom agents | — | ||
| Cost analytics (Autopilot dashboard) | — | ||
| Self-hosted agents | — | ||
| SSO (SAML / OIDC) | — | — | |
| Audit log retention | 90 days | 1 year | Up to 7 years (default 2y) |
| Multi-seat / org management | — | — | |
| SLA + dedicated support | — | Dedicated + SLA |
// existing beta users automatically get 90 days of Pro · we never see your code or data
// AGENTNETES_FOR_BUSINESS
Beyond engineering.
Whole-business autonomy.
The same orchestrator that runs your engineering agents runs your business ones. Six pre-built executive agents covering finance, legal, operations, people, marketing, and revenue.
Chief Financial Officer
- · Cash-flow modeling + burn-rate alerts
- · Monthly close automation
- · Vendor cost optimization
Chief Operating Officer
- · Process orchestration across teams
- · SLA + KPI tracking
- · Supply-chain incident response
General Counsel
- · Contract review + redlining
- · Compliance tracking (SOC2, GDPR, HIPAA)
- · IP + trademark monitoring
People Operations
- · Talent screening + onboarding
- · Performance review automation
- · Headcount planning + reporting
Chief Marketing Officer
- · Campaign orchestration across channels
- · Content production pipeline
- · Attribution + funnel analysis
Chief Revenue Officer
- · Pipeline qualification + forecasting
- · Account health monitoring
- · Renewal + expansion plays
// six agents · stable releases — Join the waitlist for access
// MULTI_TENANCY
Org → Workspace → Project.
Hard isolation at every layer.
Run one orchestrator. Serve many tenants. Per-project namespaces, per-workspace billing, per-org white-label — the structure consultancies need to ship to multiple clients without leaking state between them.
White-label at org level
- · Custom domain + TLS
- · Brand tokens (logo, colors, fonts)
- · SSO + custom auth flows
Operator console per workspace
- · Cost + token rollups per project
- · Role-based access (admin / dev / viewer)
- · Tamper-evident audit log
Isolation per project
- · Namespace-scoped agent fleet
- · Per-project encrypted vault
- · Independent etcd registry tree
Join the
waitlist
We're opening access in waves. Tell us who you are and we'll reach out when your spot is ready.
You'll get a personal email from one of the founders within 48 hours. No marketing blasts, no drip.
Engineering teams running real fleets, consultancies ready to publish agents, and operators with named compliance constraints get early slots.
You're on the list
We'll email you when your spot opens up. In the meantime, check out the documentation to see what you can build.