Actions runtime for AI agents

Give agents hands. Keep yours on the controls.

Zanii Connect is the runtime where AI agents take real action. One OAuth into every app, a registry of governed tools, a human on the moves that matter, and an audit trail for all of it.

zanii connect · runtime live actions
Connects the apps your agents already need
Connect · Act · Govern

Three jobs, one runtime.

Everything an agent needs to do useful work in the real world, without you rebuilding auth, integrations, and guardrails for the tenth time.

01Connect

One OAuth, every app. Link Gmail, Slack, HubSpot, Notion, Stripe and more through a single flow. Share a connection across your whole org, and get warned before a token silently expires.

One OAuth, every app

02Act

A registry of real actions. Browse providers and the tools your agents can call. Run them now, or queue long jobs and poll. No public API for a task? Drive a headless browser instead.

Browse the registry

03Govern

Approvals, roles, and secrets. Hold risky actions for a human, approve from a link. Scope every user with roles and permissions, keep secrets per tenant, and log every move.

See the approval flow
Human in the loop

Some actions shouldn't fire on their own.

Zanii holds the ones that matter, sends a confirmation link, and only runs on a yes. Every decision is token-authenticated, needs no login, and lands in the audit trail. Approve a wire transfer from WhatsApp in two taps.

Model Context Protocol

Already speaks Claude.

Point any MCP client at a single endpoint and every governed tool shows up as a callable function. No glue code, no bespoke tool definitions, no drift between what the agent thinks it can do and what it's allowed to do.

# add Zanii Connect to any MCP client
endpoint https://vault.zanii.agency/v1/mcp
auth Bearer $ZANII_API_KEY

# the agent now sees your tools
tools/listconnect · execute · approve · audit
tools/callruns, gated by your roles
The full surface

Built like infrastructure, not a demo.

Multi-tenant from the first request. The plumbing you'd otherwise spend a quarter building is already here.

Connection health

Per-connection status so the UI can prompt re-auth before things break.

GET /v1/connect/connections/health

Triggers & webhooks

React to events. Fire on triggers. Fan out to your own subscribers.

POST /v1/triggers

Async jobs

Queue long runs on the worker and poll a job id until it's done.

POST /v1/tools/execute_async

Per-tenant secrets

A scoped secrets vault, never shared across organizations.

PUT /v1/secrets

Roles & permissions

Define roles, grant permissions, and see exactly what a key can do.

GET /v1/auth/me

MFA & API keys

TOTP enrollment, scoped API keys, revoke in one call.

POST /v1/auth/keys

Usage metering

Track consumption per organization for billing and limits.

GET /v1/org/usage

Headless browser

When there's no API, run a real browser session and script it.

POST /v1/browser/session
On the record

Every action, accountable.

List executions, stream the whole audit trail as CSV for compliance, and honor GDPR with a one-call tenant export or hard delete.

100%
of tool runs logged with actor, input, and result
CSV
full audit trail streamed for compliance export
GDPR
portable tenant export & hard erasure on request
1 key
scopes an org's users, roles, and permissions

Ship the agent.
Skip the plumbing.

Connect once, act everywhere, and approve what matters. Get a key and point your agent at the runtime.