Your coding agents,for a fixed monthly fee
Your developers spend the day with Claude Code, Codex or Cursor, and the model’s bill arrives at the end of the month without anyone having been able to predict it. With Agent Libre behind the tool, the model answering is served in Canada and the month’s amount is written in the contract. The tool itself does not change.
Talk about your coding agentsWhat goes wrong with a model billed by usage
Three things a development team lives with when the model behind the tool is paid on demand.
You pay for the length of the session
Before each line it writes, the tool re-reads the open files, the conversation history and the test output. An afternoon spent getting a test suite to pass re-reads the same module dozens of times. Billed by usage, what is re-read costs more than what is written, and the bill follows the length of the session.
The end of the sprint shows up on the bill
A quiet week and a week of bug fixing before a release have nothing in common on the meter. The team lead finds out the amount after the fact, and cannot tell management at the start of the month what the tool will cost.
The repository leaves the country on every turn
Every call sends excerpts of the repository to the model provider: the payment module, a configuration file with its passwords, a database extract pasted into the conversation to reproduce a bug. With a foreign provider, all of it leaves the country, on every turn.
You, your tool, and the model that answers it
Your tool stays the same. What changes is the model that answers it.
Your organization
Your teams, your data and your rules. You decide what the tool does, and with which data.
Claude Code, Codex or Cursor
The tool reads your code, writes more of it and gets the tests to pass, at your developers’ request. Your team installs it on each workstation and chooses its settings.
The model that answers the tool
Every time the tool needs an answer, it calls a language model. That model can come from different providers. This is where Agent Libre comes in.
What Agent Libre changes behind your tool
The tool stays your team’s, with its commands and its extensions. Here is what changes once Agent Libre’s model is behind it.
One setting to change
Claude Code, Codex or Cursor is installed and configured by your team. It calls a language model; that model, Agent Libre serves in Canada. Connecting takes three configuration lines, in the tool’s settings, one workstation at a time.
A reserved speed, a price known in advance
You reserve a processing speed for the team. The month’s amount does not move. When the team exceeds that speed, what varies is the wait.
Re-reading the history costs nothing more
What the tool re-reads on every turn is part of the plan. Nobody has to close a conversation or trim what the tool keeps in mind to save money: a module is read in full.
Source code in Canada
The model is hosted and operated in Canada. Your code does not leave the country, is never used to train a model and is not kept after processing.
What holds for the whole team
The price, set in advance
The month’s amount is in the contract. It is revised at renewal, from measured usage, up or down.
At peak, you wait
When the whole team launches tasks at the same time, requests wait their turn. No cut-off mid-task, no extra charge.
Secrets masked on the way
An access key or a password sitting in a file sent to the model is masked by the Agent Libre gateway before it reaches the model.
What each team decides before connecting the tool
Three decisions to make before changing the model’s address. None of them touches the code.
Check that your developers can choose the model’s address in Claude Code, Codex or Cursor: it is the only setting to change. Plan a week where both models run side by side, on your own repositories and your own tasks, to compare quality before switching everyone.
Decide which repositories may go through an agent and which stay out of reach, for example the payment module or anything touching customer data. Then specify what the gateway must mask on the way: access keys, passwords, internal server addresses. The details of the guardrails are on the Governance page.
Decide whether the reserved speed covers the ordinary day or the peak day. The first costs less and makes the team wait at the end of a sprint; the second costs more and rarely waits. The free 24-hour to one-week audit, connected in front of your current provider, measures your real day, with its average and its peaks.
What we get asked on this topic
Does Agent Libre provide the coding tool?
No. Claude Code, Codex or Cursor comes from its publisher, and your team installs and configures it. Agent Libre serves the model answering behind it, for a price known in advance.
Does the model write code as well as Claude Opus?
Not on every task. The model served is GLM 5.2. On long jobs in a real project, Claude Opus keeps a small lead; for the bulk of daily work, tests, fixes, small refactorings, the gap is thin. Your own repositories, during the comparison week, settle it better than public evaluations.
What does the model see of my repository?
What the tool sends it on each turn, and nothing else: the open files, the excerpts it chose to read, the test output and the conversation. Keys and passwords are masked on the way, and nothing is kept after processing.
If the team doubles, does the bill double?
No, not within the month. The month’s amount does not move. If the reserved speed is no longer enough, the team feels it through the wait; that is the signal to revise the plan at renewal.
Do we have to switch the whole team at once?
No. The setting lives in each developer’s tool. Two or three workstations can go through Agent Libre while the others stay with the current provider, and going back means putting the old address back.
See also
Autonomous AI agents
What an agent that re-reads its history on every turn costs, and what changes when the model behind it is served in Canada for an amount known in advance.
Governance and compliance
The monthly report, the inventory of AI tools and the rule review your compliance team produces from the gateway, with no request to IT.
Pointing Claude Code at a self-hosted endpoint
The three configuration lines, step by step, and what changes in the tool once connected.
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.
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.