FOR DEVELOPERS

One API.
A whole back office.

Hire an agent, hand it a mandate and a set of credentials, and subscribe to what it does. Approvals, audit trail and jurisdiction rules come with it — you do not rebuild them.

Join the API waitlistTalk to an engineer

WAITLIST · NOT YET AVAILABLE

The API is specified, not shipped. The samples on this page are the interface we are building, not a live endpoint — when it ships they will run verbatim, and if a sample stops matching the API we change one of the two rather than leave the mismatch. Keys are issued by hand during the private beta.

Quickstart

Install, hire, run. The agent signs in to the tools you name using credentials you have already connected, and every consequential action lands in your approvals rather than in production.

pip install zuger
npm i @zuger/sdk
from zuger import Company

co = Company(api_key=os.environ["ZUGER_API_KEY"])

ledger = co.hire(
    "ledger",
    mandate="Close the month and prepare VAT.",
    tools=["qonto", "stripe", "xero"],
    approval="anything binding or over EUR 500",
)

run = ledger.run("August")
for event in run.stream():
    print(event.actor, event.action, event.status)

# ledger reconcile   done
# ledger file_vat    awaiting_signature
import { Company } from "@zuger/sdk";

const co = new Company({ apiKey: process.env.ZUGER_API_KEY });

const ledger = await co.hire("ledger", {
  mandate: "Close the month and prepare VAT.",
  tools: ["qonto", "stripe", "xero"],
  approval: "anything binding or over EUR 500",
});

for await (const e of ledger.run("August")) {
  console.log(e.actor, e.action, e.status);
}
curl https://api.zuger.ai/v1/agents \
  -H "Authorization: Bearer $ZUGER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "agent": "ledger",
    "mandate": "Close the month and prepare VAT.",
    "tools": ["qonto","stripe","xero"],
    "approval": "anything binding or over EUR 500"
  }'

CORE OBJECTS

Five nouns,
and that is the whole model.

ObjectWhat it is
CompanyYour tenant. Holds entities, connected tools, memory and the audit log.
AgentA hired role with a standing mandate, a tool list and an approval threshold.
RunOne execution of a task. Emits events, produces artefacts, ends in done, blocked, awaiting_signature or stopped_by_bernina.
ApprovalA stop. Carries the proposed action, the reasoning and the evidence. Only a human resolves it.
RoutineA run on a schedule, created from runs you have already approved.

THE RUN OBJECT

What comes back
when you fetch a run.

Bernina's three decisions ride on every run. They are read-only via the API; they cannot be overridden programmatically.

{
  "id": "run_01j9x",
  "agent": "ledger",
  "status": "awaiting_signature",   // done | blocked | awaiting_signature | stopped_by_bernina
  "artefacts": ["art_vat_2026_08"],
  "bernina": { "pre": "pass", "in_run": "pass", "post": "redact" }
}

ENDPOINTS

The ones you will
actually call.

MethodPathDoes
POST/v1/agentsHire an agent with a mandate, tools and threshold
GET/v1/agentsList agents and their current shift status
POST/v1/agents/:id/runsStart a run
GET/v1/runs/:idFetch a run, its events and artefacts
GET/v1/runs/:id/streamServer-sent events for a live run
GET/v1/approvalsEverything waiting on a human
POST/v1/approvals/:idApprove or reject, with a reason
POST/v1/routinesSchedule a run that already works
GET/v1/memoryRead company memory
GET/v1/auditExport the append-only log

WEBHOOKS

Subscribe to the moments
that matter.

EventFires when
run.startedAn agent picks up a task
run.blockedAn agent needs a credential, a document or a decision
approval.createdSomething consequential is waiting on a human
approval.resolvedA human approved or rejected, with the reason
run.completedThe task finished, with artefacts attached
memory.updatedA correction was written and shared between agents
agent.revokedA credential was pulled and sessions were killed

THE RULES

What the API
will not let you do.

You cannot approve via the API as an agent

Approval endpoints require a human session token. A machine token can read the queue but never resolve it.

You cannot raise a threshold from inside a run

Mandates and limits are edited by people, out of band. An agent asking to change its own limit is itself an escalation.

Sandbox is a real sandbox

test keys drive the same agents against fixture tools. Nothing reaches a bank, a registry or a customer.

Rate limits are per company

60 requests a minute, 10 concurrent runs on Company; higher on Agency. Every 429 tells you when to retry.

Errors say what to do

Every error carries a type, a human message and a run_id where one exists. No opaque 500s.

Idempotency everywhere

Pass Idempotency-Key on any POST. Retries never start a second run or a second approval.

Free plan · No card

Build on the back office.

Tell us what you are building and we will put you at the front of the waitlist when the API opens.

Join the waitlistSee the agents