Your goal: see all AI in your company

ARGENT FORAI VISIBILITY

Book a demo

Master Control System · by Nyedis

01 / AI VISIBILITY

Your goal: see the AI at work in your company

If your goal is to see the AI at work in your company, start from a hard fact: much of it was never registered with anyone. People use AI to draft, summarize and write code because it makes their work easier.

  • Microsoft found 78% of AI users bring their own AI tools to work.¹
  • IBM found security incidents involving unapproved ("shadow") AI more than doubled, to 43% of the breaches it studied from 20% a year earlier, at an average cost of $5.39 million.²

You cannot govern AI you cannot see. Argent gets you there in order: find it, watch it, then enforce, with everything in one console.

02 / AI VISIBILITY

Six questions visibility answers

Visibility is more than a list of tools. It is a straight answer to six questions, in one console.

QuestionWhere Argent answers it
What AI is in use, and where does it come from?The Discovery page and the Agent inventory, from your proxy logs and managed machines. Live traffic lists every request through the Argent gateway and where it came from.
Who owns each agent?Every registered agent has a named owner. The Spend page counts the dollars spent by agents with no live owner.
What sensitive data is going out?Data Loss Prevention records what it found, by kind and count, never the value, in watch mode and in enforce mode.
What does it cost, in dollars and tokens?Spend by agent, owner, model and provider, and each agent's tokens in, tokens out and cache tokens.
What would a control have stopped, and why?The Rollout page: would-refuse and would-change counts per control, with sample events and the reason.
What exactly happened in that request?Request replay: the prompt, the response and each tool call, node by node, as your capture setting keeps them (full capture is the default). See Session Replay.
03 / AI VISIBILITY

The path: from proxy logs to enforce mode

Each step adds detail, and nothing changes for your people until you decide.

  1. Import your proxy logs (Level 1). Your web proxy sends Argent a copy of the access log it already writes, in Squid or NDJSON format. Nothing is installed and no traffic is rerouted. Argent keeps only visits to AI services it recognizes, drops every other line before anything is stored, and keeps who reached which service, never what was said. See Shadow AI Discovery.
  2. Add managed machines (Level 2). Your device management (Jamf, Kandji, Intune-style or shell/PowerShell scripts) installs the Argent certificate and proxy settings once. Only machines from ranges an administrator approved are admitted, and the traffic is tunnelled, never decrypted. Anything seen that nobody registered lands in the Agent inventory tagged Discovered.
  3. Register the agents you keep (Level 3). A developer registers each agent with a named owner, so it gets an identity and a record of every request. See Identity Tethering.
  4. Watch. Registered agents start in watch mode: every request passes unchanged, and what each control would have refused or changed goes on the signed record. Console sign-in rules, revoked keys, quarantine and kill switches still apply in watch mode. See Watch Before You Enforce.
  5. Enforce, one control at a time. Admins mark sample events right or wrong and read what a switch would have refused in the last 7 days. Then they move one control to enforce with a fresh passkey check, and the others keep watching. From there Sensitive Data Redaction, Model Control, AI Spend Caps and Agent Reach Control do the stopping.
04 / AI VISIBILITY

Everything visible in one console

One console for security, operations and finance. It shows the agent, its owner and the tenant by name.

PageWhat you see
Operations (Home)A status strip for every control, a queue of what needs a person (discovered agents, agents with no owner, spend spikes, unpriced models and more) and a chart of the last three hours of requests, refusals, detections and spend.
Live trafficEvery request through the gateway as it happens, with the agent, the decision and a cost view. Click a row to inspect it, or contain the agent on the spot.
Live Flow and Activity flowWhere requests go and where they are stopped: a live flow graph and world map by organization, department, person or agent, and a 24-hour view from agent to AI service to outcome, would-refuse included.
Threat mapEvery agent on one map, with lenses for cost, request rate, budget, sensitive data, model policy, prompt injection, tool use and credentials.
Discovery and Agent inventoryEach discovery level's state (Off, Awaiting approval, No data received, Receiving data) and every client seen, tagged Discovered or Registered. A quiet feed reads "No data received", never "no AI usage".
Sessions and replayEach agent conversation with its owner and the tool it ran in, such as Claude Code or Codex, replayable turn by turn.
RolloutWhat each control would have refused or changed on your own traffic, before you enforce it.
05 / AI VISIBILITY

A live view of token spend

Every request an agent sends to a model is metered and billed. Argent counts it as it happens, in watch mode and in enforce mode alike.

  • Gartner predicts AI inference costs per agentic workflow will rise more than five times through 2028, as advanced reasoning models typically use five to 30 times more tokens than basic chatbots.³
WhereWhat it shows
Live trafficCost updates stream to the console with each request, and a cost view puts the dollars beside the traffic.
OperationsSpend in the last-three-hours chart, top spend, and budgets about to run out.
SpendDollars by company, organization, department, owner, agent, model and provider, with top spenders and a watch list. An unusual-spend check flags an agent whose spend spikes against its own 28-day median.
Agent page, Models tabEach model the agent calls, with requests, tokens in, tokens out, cache tokens, cost and cost per day.

Cost comes from the usage figures in each model's own reply, priced from a catalog by provider and model, and each signed record carries its cost. To put a ceiling on it, see AI Spend Caps and the budget and token spend goal.

06 / AI VISIBILITY

Proof you can hand an auditor

Seeing it today matters. Proving what you saw, months later, is what an audit asks for.

  • Discovery is signed. Each proxy-log window writes a signed record of who reached which AI service.
  • Every would-have-done is signed. What each control would have refused or changed goes on the signed record, and every RIGHT or WRONG label is a signed admin event.
  • Mode history is provable. Signed events show which controls were in watch and which in enforce at any moment.
  • Cost is provable. Each signed record carries the cost of its request.
  • Reports. Eight posture checks against the previous period, including agents without an owner, on one page.

What frameworks already ask for:

  • NIST AI Risk Management Framework, GOVERN 1.6: "Mechanisms are in place to inventory AI systems and are resourced according to organizational risk priorities."⁴
  • For banks and insurers supervised by the New York Department of Financial Services (NYDFS): its AI guidance adds no new requirements, and says entities that rely on AI "should maintain an inventory of all such systems."⁵

Package it for an examiner with One-Click Audit Evidence, or take it to the board with Give the Board What It Wants.

Argent Strategic Intelligence

Book a demo

See Argent Strategic Intelligence stop a real AI agent, live, with the engineers who built it.