Dashboard

The moment you enable a Connector, it gets a private cockpit. Not a settings page buried in menus: a Dashboard with eight live numbers and a feed of everything the connector just did, refreshing on its own every ten seconds. You enabled GitHub in the Library; clicking it opens this screen, and this is what a healthy connector looks like:

R

Requests 24h

3,412

THROUGHPUT · 2.4 req/min

p95 latency

412

P95_MS

Error rate

0.03%

SLA

Avg Response

187

AVG_MS

Slowest Tool

search_code

512MS

Busiest Tool

create_issue

1,204_CALLS

Active clients

3

CLIENTS_24H

Last Activity

2m ago

LAST_REQ

Live activity

14:32:07	create_issue	MUTATION	201	187ms
14:31:52	search_documents	QUERY	200	512ms
14:31:30	list_issues	QUERY	200	143ms
14:30:58	create_issue	MUTATION	201	176ms
14:30:11	get_file_contents	QUERY	200	98ms
14:29:44	create_pr_comment	MUTATION	403	88ms
14:29:02	search_code	QUERY	200	121ms
14:28:37	list_pull_requests	QUERY	200	134ms
14:28:01	create_issue	MUTATION	500	613ms
14:27:40	search_code	QUERY	200	104ms

The eight numbers

• Requests 24h counts everything this connector did today, and the card's footer shows the live pace (THROUGHPUT · 2.4 req/min). You can tell at a glance whether today is a quiet day or your agents went to work.
• p95 latency is the number your users actually feel: the slowest 5% of requests. Averages hide pain; p95 refuses to.
• Error rate wears the SLA badge, and here is a detail worth knowing: this card counts only errors Vinkius itself caused. Your agent's 4xx mistakes and the provider's 5xx hiccups are tracked too, but the SLA card is Vinkius grading itself, in red the moment it slips. Hover the number and the tooltip splits everything: agent errors, upstream errors, Vinkius errors, and the reliability percentage.
• Avg Response is the everyday speed of the connector, in milliseconds.
• Slowest Tool names the laggard in purple with its average in the footer (512MS). One name, so you know exactly what to complain about or configure.
• Busiest Tool names the workhorse in amber with its call count (1,204_CALLS). Usually this is the tool earning its keep.
• Active clients counts how many AI clients used this connector in the last 24 hours.
• Last Activity answers "is it alive?" in two words (2m ago).

Live activity: your AI, second by second

Below the numbers, the Live activity feed lists the ten most recent executions with the precision of a cockpit log: the exact time (14:32:07), the tool that ran, the kind of action (QUERY or MUTATION), the HTTP status color coded (emerald for success, amber for a refused request like a missing scope, red for a failure) and the response time.

The console refetches this feed every ten seconds on its own. You are not looking at a report about the past; you are watching your AI work.

Setup pending? The banner tells you

If a connector needs credentials and they are not configured yet, an amber banner appears above the numbers: "GitHub requires credentials. This Connector will not work until you set it up. Connect your account so AI agents can start using it." with a Set up now button that takes you straight to the Credentials tab. No silent failures: the connector tells you what it needs in the same breath as the metrics.

From the cockpit to the whole machine

The connector's sidebar keeps everything one click away, grouped the way you use it: Dashboard alone; Credentials and Connection Tokens for identity; AI Capabilities, Playground and Deployments for power; Audit Logs and Log Streaming for the record; Settings and Monetize for control and earnings. Each of them gets its own page in this sidebar, in this exact order.