Skip to content
Gate

Dashboard tour

The dashboard is where you watch traffic, tune security, and manage billing for your gateway. The sidebar groups pages into Monitor (what happened), Manage (the controls that shape requests), Gateway (the model catalog), and Workspace (account and team). This page walks through each, and calls out the ones you will use most.

Overview

Your landing page after signing in. The top row tracks message volume, tokens saved, and threats detected for the week. Below it, a tokens-used chart breaks usage down by model, and three live feeds show your latest messages, conversations, and security events.

Gate dashboard Overview page: weekly cards for messages, tokens saved, and threats detected; a tokens-used-by-model chart; and live feeds of the latest messages, conversations, and security events.
The Overview page: weekly activity, token usage by model, and live feeds.

Monitor

Everything that already happened, after the fact.

Messages

The core log. Every model message across your stack, inspected for injection, PII, and credentials before it reaches the model. Each row shows status, security result, model, token counts, latency, and cost. Filter by provider, model, status, time range, security result, or key, search the message body, or export to CSV.

Gate dashboard Messages page: a message-volume sparkline over success and error counts, above a table of recent messages showing time, status, security verdict, model, conversation, key, input and output tokens, latency, and cost.
The Messages log, with per-message status, security verdict, tokens, latency, and cost.

Open any message for the full record: the prompt and response, the security scans it passed or failed, and request metadata such as the model, provider, endpoint, HTTP status, and whether the response was a cache hit.

A single Gate message record: a Passed panel listing prompt-injection scan, PII redaction, and credential-leak detection all passing, and a Details panel with timestamp, model, provider, API key, endpoint, HTTP status, and cache result.
A single message record: its security scans and request metadata.

Conversations

A conversation is a chain of messages that share session context: an agent run, a multi-turn chat, or a tool-calling loop. Each row rolls up one session’s turns, message count, tokens, and cost; click it to read the whole thread end to end instead of scanning isolated calls. The overview tracks total conversations, average turns per conversation, and average cost per conversation over the period you select.

Gate dashboard Conversations page: overview cards for total conversations, average turns, and average cost per conversation, above a table of conversations showing each session's title, key, models, turn and message counts, tokens, cost, and last-updated time.
Conversations: each session's turns, tokens, and cost, with per-period averages up top.

Security Events

Every injection, PII, and credential event your policies caught: blocked, flagged, or redacted, each fingerprinted to Constellation’s Digital Evidence layer. The overview totals events over time and breaks them down two ways: by action (blocked, flagged, redacted) and by attack type (PII/PHI, prompt injection, credential leak). Below, the feed lists each event with its detection type, conversation, key, and action. Filter by type, key, or action, search the feed, or export it to CSV.

Gate dashboard Security Events page: a total-events time series, an action-type breakdown of blocked, flagged, and redacted counts, an attack-type breakdown of PII/PHI, prompt injection, and credential leak counts, and a feed of recent events showing time, detection type, conversation, key, and action.
Security Events: totals over time, action and attack-type breakdowns, and the event feed.

Audit Trail

A tamper-evident record of every message, response, and policy decision the gateway handled. Each event is hashed and periodically fingerprinted to Constellation’s Digital Evidence layer, so anyone can verify independently that a record has not been altered after the fact. The overview shows how many events are logged and when the last fingerprint was anchored, with a link to open it in the explorer. Each row carries an event id, type, description, the member responsible, and its fingerprint; filter, search, or export to CSV.

Gate dashboard Audit Trail page: overview cards for total audit events logged and the time of the last fingerprint with a link to open it in the explorer, above a table of recent events showing time, event id, event type, description, member, and fingerprint hash.
The Audit Trail: fingerprinted events with a link to verify them in the explorer.

Manage

The controls that shape how requests are handled.

Policies

The three inline scans that run on every routed message: prompt-injection detection, PII/PHI redaction, and credential and secret leak detection. Turn each on or off, tune its sensitivity and what it detects, and choose how a hit is handled: flag it, redact it, or block the request. A message’s detail view shows the result of each scan.

Gate dashboard Policies page: three inline scans, each shown ON with an expandable settings row: prompt-injection detection, PII/PHI scanner, and credential and secrets scanner.
Policies: the three inline scans, each with its own sensitivity and action.

Limits

Spend and rate limits, attached to a key or to the whole organization. Set ceilings so a runaway client cannot surprise you.

Token Savings

Caching, compression, and de-duplication, with a running tally of what they saved over the selected period. Turn on response caching (with a TTL) so identical messages are served without a round trip, and enable compression to trim input tokens before they reach the provider.

Gate dashboard Token Savings page: overview cards for total saved, caching, and compression with trend lines, above toggles for compression (including advanced options) and response caching with a TTL setting.
Token Savings: caching and compression controls, with savings tracked over time.

Gateway

Models

The catalog of every model Gate can route to across your enabled providers, with per-provider input and output pricing, context window, and capabilities. Each model has a detail page with code samples. This is where you find the provider/model id to put in your requests, and where owners and admins enable models or pin a preferred provider when more than one can serve a model.

Gate dashboard Models catalog: a searchable table of models across four providers, each row showing the model name, its provider/model id, context window, input and output price per million tokens, capabilities, and which providers serve it.
The Models catalog: pricing, context, and capabilities across every provider.

Workspace

Account and team administration.

Activity

A friendly feed of who changed what: security decisions, key creation and rotation, billing transactions, and member changes.

Team

The members of the active organization, their roles, and pending invitations. Invite a teammate, change a role, or remove a member. See Teams and roles.

Billing

Your plan, prepaid balance, and payment method. Top up here; changes open Stripe Checkout or the customer portal. See Plans.

API Keys

Create, name, rotate, and disable your Gate keys (sk-gw-…). Each key can carry its own usage limits and security policy. See Authentication for how keys authenticate requests.

Settings

Per-organization configuration: display name, environments (tags for dev, staging, and prod), default provider account, and default security policy.