Tender season lands at the same time every year: five hundred-page bids to compare in ten days, with a purchasing department already busy. The tool your provider or your team puts in place reads those documents and calls a language model to do it. With Agent Libre behind it, that model answers from Canada, for a fixed monthly amount.
Talk about your supply chainThree situations procurement and logistics teams live with.
One supplier includes delivery in the unit price, another bills it separately, a third indexes its price from the second year on. The buyer puts everything on one grid by hand, and the grid takes three days. Your provider’s tool produces that grid from the complete documents, if the model behind it reads each bid in full.
Late penalties, warranties and price-revision terms in a three-year contract are written down, often in an appendix. Nobody re-reads them until a delivery is late. Your team needs a tool that pulls them out of the contract at signing and lays them out in a table, with the page they came from.
At receiving, quantities, part numbers and sometimes prices differ from one document to the next. The gap turns up at month end, when the invoice arrives. The tool matches the three documents at the moment of receipt; that is reading work repeated hundreds of times a month.
Your tool stays the same. What changes is the model that answers it.
Your teams, your data and your rules. You decide what the tool does, and with which data.
The tool that compares your bids, pulls the clauses out of your contracts and matches delivery slips to purchase orders. Your provider or your team installs it and connects it to your purchasing system.
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.
The tool that reads your documents calls a language model. That model can come from Anthropic, from OpenAI or from Agent Libre. Agent Libre serves that model. The tool itself stays your vendor’s or your team’s.
Your provider or your team puts in place the tool that receives documents from your purchasing system or your ERP and sends its grids and tables back there. It has to let you choose the address of the model it calls. The validation rules, what the tool decides alone and what a person approves, stay in that tool.
To compare five bids, the tool re-reads the specifications five times. To pull out clauses, it re-reads the contract at every question. With Agent Libre behind it, that re-reading is not billed per token: it counts in the sizing of your PTU, and the cache keeps the specifications already read, which makes each re-read about five times lighter. The month’s amount stays fixed.
You reserve a number of PTU (provisioned throughput units) for your tool, sized on your peaks, counting tokens read as well as tokens written. During tender season, documents go through at that throughput. At renewal, the plan is revised from the measured volume, up or down.
The model is hosted and operated in Canada. A public tender, a prime contractor’s technical documentation or a subcontracting quote does not leave the country and is never used to train a model. Agent Libre’s gateway masks personal information as it passes through (signatories, contact details, account numbers) and logs every call.
The specifications, their appendices and each bid are read in one pass. The tool compares prices, lead times and exclusions on the complete documents, without splitting them.
The free 24-hour to one-week audit places the gateway in front of your current provider and measures what your tool consumes on your own documents. The plan is set from that measurement.
Every call to the model is logged: which tool, what time, what volume. The audit logs are kept for twelve months and serve to answer a prime contractor or an auditor.
Four decisions to make inside your organization. The model does not make them for you.
Pick a tender already awarded, one your buyers graded by hand. Have the tool re-read it with the model served by Agent Libre and compare the two grids line by line. It is the shortest test of quality on your own documents before switching.
Decide which gaps between purchase order, delivery slip and invoice the tool may settle alone, for instance a quantity difference under a threshold you set, and which ones wait for a person. That rule is written into your provider’s tool before the model behind it changes.
Reread the clause in your contracts that governs processing data outside the country and what it requires beyond server location. Agent Libre documents where the model answers, what the gateway masks and what the logs keep; the decision to run a given contract through it rests with you and your contracting authority.
Check that the tool lets you change the model’s address, the key and the model name: those are the three lines of configuration to edit. Also ask your provider where the tool keeps its own copies of documents and grids: the model keeps nothing, but the tool archives according to your rules.
Yes for location: the model is hosted and operated in Canada, and no data is sent to a foreign provider. Agent Libre documents where the model answers, what the gateway masks and what the logs keep. Your compliance team or your contracting authority judges whether that meets the clause of a given contract.
No. Your provider or your team puts the tool in place and connects it to your purchasing system or your ERP. Agent Libre serves the model that tool calls. What changes with Agent Libre behind it: the country the model answers from, and a fixed monthly amount instead of an invoice that follows volume.
On reading and extraction in long documents, GLM 5.2 stands close to the Anthropic and OpenAI models, which keep a small lead on the longest tasks. The test described above, a tender already awarded re-read by both models, settles it for your documents better than any public evaluation.
Documents go through at the throughput of your PTU (provisioned throughput units). Beyond that, requests go through on a best-effort basis, without a throughput guarantee but with no cut-off and no charge; a comparison grid may take longer to come out, and the month’s amount does not move. If your peaks are known, the number of PTU is chosen with them counted in, not only from the average.
It sees the text the tool sends it: specifications, bids, contracts, converted to text. It does not see drawings or images. Before the text reaches the model, Agent Libre’s gateway masks what you tell it to, for instance the names of signatories or suppliers’ banking details. Nothing is kept beyond processing.
What the agent connected to the ERP asks of the model that serves it, and what finance, purchasing, payroll and IT check before changing that model.
What a ministry, a city, a hospital or a college decides before putting a model served in Canada behind its tools, and what the auditor finds there.
Why a reserved processing speed suits a document volume that changes from one month to the next.
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.
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.