Inbox
Everything that needs a person, in one place: requests waiting for approval and alerts from every part of Vane.
Investigate
Start from any number and narrow it down until you see the requests behind it.
What stands out found automatically · check before acting
Per day
Break it down by
The requests metadata only · never the text
Executive overview
Requests per day last 14 days
Spend
Sensitive data seen
Agents
| Agent | Requests | Cost | Owner |
|---|
Recent requests
| Time | Agent | Project | Model | Tokens in / out | Cost | Overhead | Decision |
|---|
Policies
What the gateway allows, changes, holds or blocks, checked on every request.
Start from a template
Change history
| When | Policy | Change |
|---|
Approvals
Requests held by a policy. Each one goes to the requester's managers; approving lets that agent through for a set time, and it retries on its own.
Decided
| Decided | Agent | Requested by | Policy | Outcome | Decided by | Note |
|---|
Activity
Every request a policy acted on, or would have in monitor mode.
| Time | Agent | Project | Decision | Policies | Why |
|---|
Agent map
Agents running on the organisation's network, who made them, where they connect and when. Updated continuously.
What each agent touches approved path sensitive, check it not allowed numbers = connections
Who connects to what
Agents
| Agent | Status | Made by | Reaches | Last connection |
|---|
Recent changes
Sources
| Source | Last read | Items | Problem |
|---|
Models
Every AI provider and model in the world catalog. Only the ones the company approves can be used: the gateway refuses the rest, and firewalls block their sites with the list below.
New from the weekly check (blocked until approved)
Weekly vendor check
| Vendor | API | Checked at | Answer | Time | Models listed | No longer listed |
|---|
Requested, but not in the catalog
| Model | Through | Requests · 30 days | Blocked | Last |
|---|
Providers and models
For firewalls and web proxies
The sites and APIs of every provider that is not approved, one domain per line, kept current:
. Use it as an External Dynamic List (Palo Alto Networks, Fortinet, Check Point and others) or a
URL-filtering list, with a rule that denies it. Read it with a viewer token (Bearer, or HTTP Basic with the token as the password).
This blocks use outside the gateway, such as employees on ChatGPT or Gemini in a browser.
System log
Every sign-in, change, decision, sync, report and export, in a tamper-evident chain. Newest first.
| When | Category | What happened | Who | Target | From | Details |
|---|
Access
Give each person the smallest role that lets them do their job. Every change is kept in the system log.
Roles ✓ yes · “team” only people below them · “own” only their own agents
Mobile app who may pair a phone, and for how long
Paired phones
| Person | Phone | Paired | Signed in until | Last seen |
|---|
Accounts
| Person | Role | Status | Last sign-in | Directory |
|---|
Access tokens
| Name | Role | Status | Last used | Expires |
|---|
Directory
People and reporting lines from your HR system and Active Directory. They decide who approves what.
How approval authority works
People
| Person | Title | Department | Reports | Status | Sources |
|---|
Sync history
| When | Result | Started by | People | Added | Left | Manager changes | Notes |
|---|
Marketplace
Every agent running in the company, with what it does, who owns it and how it is used. Find the one you need and ask its owner for access.
Waiting for your decision
| Person | Agent | Why | Asked |
|---|
Your access
| Agent | Status | Until | Calls |
|---|
Compliance · SOC 2
How ready the platform is for a SOC 2 audit: its own controls, checked live, and a scan of its code. Open any criterion to see what is behind it.
Readiness by day
Open findings
SOC 2 criteria
Findings
| Severity | Finding | Where | SOC 2 | Found |
|---|
Checks
| When | Result | Readiness | Code | Findings | Started by |
|---|
System check
Checks that the platform works, that nothing was changed without authorisation, and that no suspicious code was added.
History
| When | Result | Critical | Warnings | OK | Version | Started by |
|---|
Cost savings
Three ways to pay less for the same work. Each starts in Monitor: it changes nothing and shows what it would have done and saved. Switch a feature on for everyone, or only for some teams, agents or people.
For the whole organisation
Exceptions a team, an agent or a person that differs from the organisation
The most specific setting wins: a person, then an agent, then a team, then the organisation. A locked setting wins over everything below it. “Inherit” takes the level above.
What applies to…
Prompt caching: where the money is last 30 days
Settings admins only
A 1-hour cache costs more to write (about 2× instead of 1.25× the normal price), so it pays only when the same start repeats over a longer time.
Reports
Who gets what, and how. Each person gets the reports their role allows, as often and in the format they like.
A few pages to read or forward: numbers, charts and what needs a decision.
Up to 2 minutes, with captions, so it works without sound.
A 3 to 5 minute spoken summary for the drive to work. Arrives as an audio file and in the app.
Subscriptions a report shows only what that person may see
| Who | Report | How often | Format | Delivered to | Last sent |
|---|
Your reports also in the mobile app
| When | Report | Status | Open |
|---|
Who may receive which report
This browser cannot play the video here. Download it and open it with any video player.
Your organisation · weekly
| Week | People | Using AI | Requests | Spend | Stopped | Status | Report |
|---|
Every manager · weekly
| Manager | Week | People | Requests | Spend | Status | Sent by | Report |
|---|
Organisation · daily
| Day | Status | Sent by | Report |
|---|
