v1 built · early access opening soon

The operating system for your AI workforce.

Agent Manager HQ provisions, monitors, and governs fleets of AI agents — deployed across WhatsApp, Telegram, Slack, and more — the way you'd manage a workforce, not a pile of scripts. Signed config, cost metering, and automatic rollback, built in from day one.

No spam. One email when access opens.

fleet — 4 agents
support-agent-01
healthy v14
sales-agent-03
healthy v14
ops-agent-02
drift v13
finance-agent-01
healthy v14

Platform

Run agents like a fleet, not a folder of scripts.

As AI agents move from one-off scripts to always-on workers, the hard part stops being "build an agent" and becomes provisioning, identity, cost, and lifecycle — at fleet scale.

Provisioning

Spin up an isolated, single-tenant agent instance per team in minutes — no shared state, no noisy neighbors.

Multi-channel

One agent, deployed across WhatsApp, Telegram, Slack, Discord, and more — same governance, every surface.

Cost control

Every model call routes through a governed gateway. See exactly what each agent costs, per day, per team.

Safe by default

Config is signed and validated before it ever reaches an agent — reject on bad config, roll back to last-known-good automatically.

Fleet visibility

Health, version, and config drift for every agent you run, in one dashboard — not scattered across logs.

Secrets vault

Provider keys and channel credentials are envelope-encrypted and never touch the agent runtime directly.

How it works

A control loop, not a deploy script.

HQ never pushes into your agents — it can't, they're not reachable inbound. Instead it runs a device-shadow loop: HQ owns desired state, each agent pulls, verifies, and reports back.

01

Declare desired state

Define what an agent should be running — model, channels, permissions. Versioned and signed the moment you save it.

02

Agent pulls & reconciles

Each agent polls for its config, verifies the signature, and applies it — validating before it ever commits to a bad state.

03

HQ tracks drift

The agent reports what it actually applied. Anything out of sync — a stuck rollout, a manual edit — shows up as drift, instantly.

Where we are

Built for reliability, not vibes.

We're in active development. The core control loop — provisioning, signed config, drift detection, rollback — is built and tested end-to-end. We're opening access gradually, starting with a small group of early design partners.

284
tests, CI green
v1
thin skeleton, built
E2E
verified on a real cluster
0
inbound ports on any agent

Be first to run your fleet on HQ.

Early access opens in small batches. Join the waitlist and we'll reach out when there's room for your team.