On June 3, 2026, EightX Labs opened agnt8x to the public — an AI agent workforce platform that lets enterprises hire, onboard, manage, and orchestrate AI agents across every major LLM under one Passport, one audit trail, and one contract. The smol.ai newsletter flagged it as the most concrete shift yet from single-model thinking to multi-agent orchestration at the business layer. This guide walks through the five modules, what CONDUCTOR changes for your automation stack, and how a Halmob-style mobile and n8n team should respond.
Most teams already run agents — a Claude desk research bot, a GPT invoice triager, a Gemini code reviewer. Each lives in its own SaaS, with its own billing page, its own audit log, and its own idea of "who is allowed to do what". agnt8x argues the obvious fix: treat agents as a workforce, not as a pile of API keys. One identity (Passport), one performance ledger, one orchestrator on top.
The 30-Second Version
What agnt8x Actually Is
EightX Labs calls agnt8x the world's first AI agent recruitment and workforce management platform. Strip the marketing and what remains is a neutral control plane: it does not train models, it does not own a foundation model, and it does not lock you into one provider. It sits one layer above Claude, GPT, Gemini, and the open-weight pool, and tracks who did what, when, for how much.
The platform ships in three deployment modes:
- SaaS — multi-tenant for start-ups through mid-market.
- Tenant Workspace — single-tenant for larger enterprises that need isolation.
- EMBASSY — the full platform inside the customer's own VPC with zero agnt8x access to runtime data, for banks, regulators, and government.
That last tier is the interesting one. It is the first multi-agent orchestration product we have seen built primarily for "the data never leaves my VPC" buyers, which is exactly the bar most European and Turkish enterprises actually require.
The Five Modules in Plain Language
agnt8x splits the lifecycle of an AI agent into the same stages a company already uses for human employees. Each module solves one stage.
| Module | Human HR analogue | What it actually does |
|---|---|---|
| FIND | Job board | Ontological marketplace matching available agents to a role you describe. |
| FORGE | Internal mobility | Private enterprise catalogue for your own agents and skills. |
| STUDIO | Onboarding flow | Nine-step setup — permissions, tools, memory, evals, escalation. |
| MANAGE | Performance & payroll | Real-time P&L per agent, alignment monitoring, unified billing. |
| CONDUCTOR | Team lead | Orchestrates multi-provider agent teams on a single canvas. |
The MANAGE numbers are the part most teams underestimate. Today, nobody actually knows what an individual agent costs them per completed task, because the billing line at OpenAI or Anthropic bundles every product together. A per-agent ledger turns "AI spend" from a black box into a unit-economics question you can answer.
Why CONDUCTOR Is the Headline
CONDUCTOR is the module that puts agnt8x on the orchestration map. It runs multi-agent teams from different providers on one canvas, in the same workflow, with one set of permissions and one audit trail. EightX claims no other platform does this yet — and while "first ever" claims age badly, the cross-provider canvas with shared identity is genuinely rare.
We have written before about the broader shift from one big model to a small router plus a plural worker pool in Sakana's Conductor and the 7B router pattern. CONDUCTOR sits at the business-process layer of that same idea: Sakana routes per subtask inside one workflow, agnt8x routes per agent across many workflows.
The shift from agent SaaS to an agent workforce is the same shift the cloud made from rented servers to managed services. The unit of buying moves up one level.
How This Changes Your Automation Stack
If you already run n8n in production, or if you operate a Hermes / deepagents harness for coding agents, agnt8x is not a replacement — it is the layer above. The practical change for a small, operations-heavy team is concrete:
- Stop registering every agent inside every SaaS one at a time. Onboard once in STUDIO, expose via Passport, reuse everywhere.
- Stop reconciling four invoices to know your real AI gross margin. MANAGE attributes cost to the agent and the task that triggered it.
- Stop hard-coding the model name in your n8n nodes. CONDUCTOR decides which provider runs each step and you keep the rest of the flow intact.
- Stop writing a separate audit trail per tool. One Passport produces one timeline that compliance can actually read.
That "onboard once, reuse everywhere" pattern is closer to the workspace agent direction OpenAI pushed earlier this year — see our breakdown of ChatGPT Workspace Agents for the in-product version of the same idea.
The Mobile Angle
The Halmob lens on this is straightforward: a workforce platform only feels real when an operator can approve, pause, or fire an agent from a phone. Web canvases impress demo audiences. Phone push approvals run actual businesses.
A CONDUCTOR-style orchestration plane plus a thin mobile approval surface is the pattern we already prototyped in Hermes Workspace mobile agent orchestration. agnt8x makes that pattern viable for buyers who refuse to install ten separate vendor apps on their team's phones — one Passport means one notification stream and one approval screen.
Why Mobile Matters for Agent Workforces
Risks and Open Questions
agnt8x is a launch, not a verdict. A few honest things to watch as the platform matures.
Vendor neutrality has to be proven, not declared
The pitch is "every major LLM". The practical question is which providers are first-class on day one, which ones lag a release, and whether tool semantics actually map cleanly between Anthropic, OpenAI, and Google. Cross-provider canvases are easy to draw and hard to ship.
The Passport is a single point of trust
Unified identity is the feature. It is also the blast radius. The first serious incident question agnt8x will face is what happens when a Passport leaks, and how granular the per-tool, per-data-source scopes really are.
Marketplaces need supply
FIND only matters if there are quality agents on the other side of it. The builder revenue share is recurring monthly, which is the right incentive, but two-sided marketplaces take 12–18 months to find out if they tip.
How to Get Started Pragmatically
- 1Read the launch material end to end — the GlobeNewswire press release covers the modules in EightX's own words.
- 2Pick one workflow you already run that calls more than one provider. That is your CONDUCTOR pilot.
- 3Onboard exactly two agents in STUDIO — one from each provider you currently pay. The point is to feel the unified-billing claim.
- 4Plug the workflow output into your existing reporting. If MANAGE's P&L line matches your finance team's number for the same month, the platform is buying you a real ledger.
- 5Only then evaluate FIND. Marketplace agents are interesting, but they are the easiest part to swap out later.
If you are still mapping out the broader agent landscape and where a workforce layer sits inside it, our OpenClaw 101 guide for new users covers the underlying building blocks — tools, skills, permissions, memory. agnt8x assumes those exist. The guide is the floor; agnt8x is the ceiling.
The Bottom Line
agnt8x is the cleanest argument yet that 2026's interesting product surface is not a smarter model. It is a workforce layer over the models you already use. CONDUCTOR turns multi-provider orchestration into a normal product feature; MANAGE turns AI spend into a number a CFO can sign off on; STUDIO turns agent onboarding into a checklist.
At Halmob we already pair n8n automation with mobile-native approval surfaces and OpenClaw consultancy for teams that want one stack to own the whole agent lifecycle. agnt8x slots into that picture as the workforce ledger — and the question to take into your next sprint is simple. If your AI agents were on payroll, would your current tooling pass an audit?
For the original coverage, see the EightX Labs announcement on GlobeNewswire's agnt8x launch announcement and the agent-orchestration recap on the smol.ai AINews newsletter.