Managed ANS
A governance layer for teams running fleets of unattended agents, built on the Tokonomix gateway. Not available yet — this page describes what's being built, not what you can buy today.
What it is
The open-source harness governs one agent, one repository, one run. Managed ANS is the layer above that: a way to point a whole fleet of ANS runs — across repos, across machines, across a team — at one Tokonomix gateway token, so the things a team needs to control centrally don't have to be configured by hand on every machine that runs an agent overnight.
The five capabilities
One token per fleet
A single Tokonomix token authenticates every unattended run in the fleet, resolved as a token-reference at launch — never a raw provider key copied into a config file on each machine.
Hard monthly budget cap
A cap enforced at the gateway, not trusted to each machine's local config. A fleet of agents that runs overnight can't quietly exceed what was budgeted for the month.
Model and provider allow-lists
Which models and providers an unattended run is allowed to call, set centrally once, rather than as a convention every repo has to remember to follow.
EU data-residency routing
Route calls through EU-based infrastructure. This is data-minimisation routing, not a compliance or certification claim — it controls where a call goes, not your regulatory status.
Central billing and usage
One place to see what every unattended run across the fleet spent and called, instead of reconstructing it from separate provider dashboards per machine.
Why this matters for a fleet, not a single run
One unattended agent on one repo is easy to reason about: you set a key, you watch the morning report, you move on. The problem shows up at the second and third repo, the second and third machine. A budget cap set locally on one box says nothing about what the other four are spending. A model allow-list remembered as a convention in one .env file isn't enforced anywhere. An EU-residency requirement that applies to the whole team can't be a per-machine setting someone might forget to copy. The mechanism Managed ANS adds is centralizing exactly these controls — identity, spend, routing, residency, visibility — at the one place all the fleet's calls already pass through: the gateway.
The honest split
What's free stays free
- The open-source harness is complete today, on your own keys. Everything in the GitHub repo — the state machine, the deterministic gates, git-backed reversibility, the ASK/PARK/HALT contract, the watchdog — works fully without a Tokonomix token.
- The paid layer is governance, not features. Managed ANS doesn't unlock capability the OSS harness lacks; it centralizes control across a fleet of runs that would otherwise be configured machine by machine. Nothing about running a single ANS instance on your own keys is held back.
- No waitlist theatre. This page is a status note, not a signup funnel. There's no email capture pretending to be a queue.
- Progress lands in the repo first. As pieces of this are built, they show up as commits and CHANGELOG entries in the GitHub repository before they show up as a marketing claim here.
In the meantime
Use the open-source harness today
Read the README → Tokonomix gateway → Questions? Read the FAQ →