🍁 Proudly Canadian — hosted and operated in Canada
White paper The digital supply chain Get the white paper
Agent Libre
Governance

Control what happens.
Know what it costs

Two questions land at the same time once AI reaches production : what is leaving your walls, and what is it costing. Both are settled in the same place — the point the traffic passes through.

Talk about your governance
The point of passage

One place to look, one place to decide

Your AI agents keep calling what they already call. The traffic runs through a point you control : that is where you see, where you block, and where you count.

Your applications
Anthropic SDK
OpenAI SDK
n8n / LangGraph
In-house AI agents
The point of passage
Inventory
Who calls what, with which model
Guardrails
Block or mask, in flight
Quotas and budgets
Per team, per application, per model
Audit logs
Full trail, self-serve reports
The models
Agent Libre — flat-rate throughput
Public providers
Self-hosted open models
Section 1 — Risk

What leaves your walls without anyone seeing it

Six problems that show up the moment AI moves past the trial stage. None of them is solved by an internal policy : you have to control the traffic.

What gets sent to the model, call by call
System prompt
10 %
Repository files•
32 %
Conversation history
26 %
Client data•
20 %
Configuration secrets•
12 %
The answer returned
• Sensitive dataThe input is the bulk of the volume — and that is where your data sits. With no point of passage, nobody knows what it holds.
01
Shadow or indirect AI

Teams wire up AI tools without going through IT. Nobody holds the list of the applications, models and providers actually in use.

02
Data and IP leaks

The context sent to the model is often more sensitive than the answer : code repositories, client files, configuration secrets.

03
Prompt injection

An instruction hidden in a document or a web page hijacks the agent. It runs someone else’s order with your access.

04
Denial of wallet

A loop that runs away, or a third party exploiting it, burns a quarter’s budget in a few hours.

05
Keys and tokens in the wild

Keys copied into scripts, shared between teams, never rotated. One leak opens everything.

06
Data reused for training

Some providers reserve the right to use what you send to improve their models. The rule sometimes changes from one version of the terms to the next, without anyone realizing.

Section 2 — Controls

What we put in place

Nothing to install on workstations, nothing to rewrite in your applications. The traffic is intercepted on the way through.

See
A complete inventory

Every call is tied to an application, a team and a model, by reading the metadata. Shadow AI surfaces on its own.

Block
Guardrails applied in real time

Ready-made rules for personal information, card numbers, secrets and sensitive code. You choose : block it, or mask it and let it through.

Prove
Compliance and logs

Quebec’s Law 25, GDPR, PCI DSS, HIPAA depending on your sector. Full audit logs and reports your teams pull themselves. The compliance team’s work is described on the Governance and compliance page.

Keep it here
Data sovereignty

The traffic never leaves Canada. Agent Libre’s infrastructure is hosted and operated here.

One rule, the way you set it
New guardrail
Security
Operational
Regulatory
Rule
Personal informationDefault
Secrets and keys
Source code
Applies to
All traffic
Mode
Mask and let through
Alert
Security team
Apply
Before
"email": "julien@example.ca",
"card": "4539 1488 0343 6467"
After
"email": "[masked]",
"card": "[masked]"
You pick the rule and the mode. The rest happens on the way through, without touching the application that calls.
The report the auditor receives
Compliance — 1 to 31 August 2026
Export
87 %
Overall
78 %
Security
100 %
Sovereignty
94 %
Logging
By requirement
Law 25 — personal information
92 % · Compliant
Masking of secrets and keys
78 % · Needs work
Data sovereignty (Canada)
100 % · Compliant
Log retention (12 months)
94 % · Compliant
Per-team access review
61 % · Overdue
12,480 calls analysed · CSV and PDF export
Your teams pull the report themselves, for whatever period they choose.
Section 3 — Cost

Three reasons the AI bill gets away from everyone

Complexity

Input tokens, output tokens, cached tokens. Models with ten different rates. A price that changes without notice.

Opacity

Millions of calls a month, and nothing tying them to a project, a team or a client.

Unpredictability

Usage that swings wildly. The bill varies by 300 % month to month, and nobody can commit an annual budget on that.

Section 4 — Taking it back

What puts the bill back under control

The same amount, two readings
What the bill shows
$38,400
a month, one line
What is inside it
Product team
$15,700
Support agents
$10,400
Data
$7,300
Trials and R&D
$5,000
The breakdown says what to cut and what to keep.
01
Attribution

Every call tied to a team, an application, a person. Without it, no useful conversation about cost is possible.

02
Alerts and forecasting

The trend across history, real-time alerts on spikes and quota breaches.

03
Quotas and budgets

Per model, per provider, per team, per application. Knowledge you can rely on.

04
The flat rate

The best cost governance is still a cost that does not move. With Agent Libre you reserve throughput : the bill is fixed.

What wakes you before the bill does
Alerts — last 14 days
J-14▲ SpikeJ-1
Quota breach
Product team at 128 % of its monthly budget
2 h ago
Spike detected
Support agents, 3.4× the 7-day average
yesterday
Forecast
At the current pace : $41,200 by month end, up 7 %
this morning
What changes

Risk and cost are settled in the same place

These two efforts are often handed to two teams, with two tools. They are looking at the same traffic.

FOR SECURITY

A real list of what is in use, rules that apply on the way through, and a trail for the auditor.

FOR FINANCE

A cost tied to whoever spends it, an alert before the surprise, and a number that holds for twelve months.

FOR THE TEAMS

No code changes, no per-seat quotas, no reason to throttle agents to contain the bill.

Where to start

The 24-hour to one-week audit

We measure your actual consumption at your current provider : who calls what, what it costs, and what is leaving your walls. Free, no commitment, and the report is yours even if you go no further.

Book the audit Write to us See the solutions
White paper

Your company has two supply chains

The first one you already know. Materials come in, they are transported, stored, transformed, delivered. Every step has a supplier, a lead time, a unit cost. When the margin tightens, you walk down that chain step by step, and you generally know where to look.

The second one moves information instead of materials. A price request comes in and becomes a quote. An order comes in and becomes a production order, an invoice, a confirmed delivery. A site report becomes a change order, or it does not. This chain runs through every one of your departments and it employs a large share of your people.