Mistral Patent for “Code implemented tool calls”

Having my name on a bunch of software patents - and, yes, I tried to get my name off them, but was not allowed - I can fairly confidently say: There is not A single worthy software patent out there. You know, one that is "not obvious to someone skilled in the art" and that actually protects a monetary investment.

Software patent are a scourge of the software industry. Patents are designed to protect costly research; simply having an idea is not costly (but it makes in medical research for example). All that software patents do is creating a minefield that hinders competition.

For software Copyright is a far better instrument. Let the one with best implementation win... That's where the cost is: Implementing, testing, shipping, maintaining. Protect that.

Sorry for the rant.

Edit: Spelling

It is quite funny an EU company patenting a software feature that is basically unpatentable in EU in the US.

Clearly this is an attempt to prevent similar patents from being weaponised against them in the US.

No one cares about such stuff in Europe.

So anyone have prior art we can point to? Because no way is a rpc call novel.

Is “by an LLM” the new “on a computer” for lame patents?

Patent for the most basic pattern seen in literally all of software engineering

This is called CodeAct (https://arxiv.org/abs/2402.01030) and is available in a number of frameworks.

Patent was applied for 04.03.2026. langgraph-codeact _shipped_ this capability in February.

It should be embarrassing and shameful to apply for such a trivial patent.

Do software patents only exist for rich players to build a moat against poorer players? At this point with patents of this quality only reason for them to exist seems to be to cross-license them for extra revenue and to possibly block anyone too poor to fight for themselves...

Setting aside the tech, which looks like an asynchronous await across a network, the interesting question is why Mistral would patent this in the US.

Simple, because the patent is less about protecting an invention than doing business in the US.

The US is more hospitable to software patents than Europe. A US patent also gives Mistral something useful when dealing with companies that already hold large patent portfolios: defensive leverage, cross-licensing, and one more asset on the books to make it attractive to investors.

There's a certain absurdity to it, and the French have a history with absurd. A French open-weights company crosses the Atlantic to put a patent around a loop, not because the code is remarkable, but because the American market makes the loop a valuable asset, business-wise.

Some confusion in this thread... this is not a patent application. This is a fully granted patent. Looks like they fast-tracked it. Filed application in March, granted in June.

You can see the actual claims here: https://ibb.co/k2mVVkFN

You can find the full patent on the USPTO search site

https://ppubs.uspto.gov/pubwebapp/

search: "12670045".did

How is the US patent registry easy to search ? I would be curious to see the patent registration of OpenAI, Anthropic, Google, etc. One way or another, I do fully expect every of those companies to try to fill these kinds of "almost for sure invalid and overturnable" patent as a way to defend themselves from every other actor (the classic "Oh I infringed on your patent ? Well here are some of MY patent you are infringing on, so let's call it a day").

Oh wow :D Not commenting on the larger picture of why or why not Mistral might be doing that, but commenting on the idea itself: I plan to build my own harness soon, and one of its core tenants was, from the beginning, that it would parse every tool call the LLM attempts and executes them itself. I most cases, that would mean to delegate to one of the tools of the rust coreutils, or making a call to a LSP server.

My reasoning: - That way I will have maximum control over how the LLM can interact with the world (i.e. I can even narrow down which arguments I support/not support) - The monitoring story should be great as well, since every interaction with the outside world happens at one place only - I can add functionality as I go, and can understand how the harness changes and why it does what it does from the ground up

The biggest challenge I foresee is with the LLMs making tool calls and having them react to the results. I also wonder how far I will even get with this :D

Back to topic; "code implemented tool calls" is basically my "oh yeah my harness can do tool calls too" (and everything I need is implemented in one codebase, in... code)

Can there be consequences for those who grant those patents? It seems the primary examiner Brian W Wathen also granted prior Art patents to Salesforce and NCR Voyix as well.

For anyone, like me, who is/was confused by the patent diagram and labeling, the initial arrows labelled "tool call execution request" (client->server) and "tool calls" (server->LLM) seem to be confusingly labelled. This isn't a tool call being sent from client, via server, to LLM (which is not a thing - tool calls go in the opposite direction), but rather the user/agent's prompt, which as always includes tool "declarations", being sent from agent, via server, to the LLM.

The big picture here is that overall tool call flow works as normal, but rather than the LLM just generating tool calls, it is instead generating a "code block" that may include multiple tool calls/invocations. The server (between LLM and client) receives this code block and will execute it, but first has to execute any contained tool calls, some which it may do locally, and some via calls back to the client/agent as normal. Once all tools calls and the containing code block have been executed by the server (in a sandbox), then the final result is sent back to the client.

Cloudflare introduced Code Mode far earlier (https://blog.cloudflare.com/code-mode/)

How does this make sense? It is public knowledge/pattern.

Given I wrote an article describing the same things years before they submitted the patent, could I not invalidate their patent with that?

I do wonder if we need a charitable foundation that does software patents and makes them freely available just to stop cockroaches abusing patents.

Another step (although governance would need to be super good for this). Would be for this charity to withold access to these patents from companies that decide to abuse stupid patents.

Software patents are the same problem as letting Billionairs buy elections and influence policy. Its stupid.

The Scala[1] community has been working[2] on something similar for a couple of years now, with papers already published[3]. I wonder if Mistral is not just trying to steal this work.

[1] https://martinodersky.substack.com/p/tracked-capabilities-fo...

[2] https://github.com/lampepfl/tacit

[3] https://www.caisconf.org/program/2026/papers/securing-agents...

Regular PSA for patent-related discussions: what matters is the claims, so before commenting one should read those.

Also this is a patent application, so it has not been examined yet. The claims at this point will always be over-broad, because that's just how patent prosecution works: the applicant (brazenly) tries to stake out as much "intellectual territory" as possible (because why not!) but the patent authorities will push back to the extent it does not overlap with other existing "territories" i.e. prior art. The applicant then narrows their claims and the back-and-forth continues until the authority determines that the claims are indeed novel and non-obvious enough to be granted. Every alternate back-and-forth costs $$ and time, which incentivizes applicants to make reasonable claims.

A typical way this negotiation often proceeds is that the applicant starts with a couple of the broadest claims they think reasonable and add multiple narrowing claims (or "dependent" claims) that cover narrower and narrower functionalities, essentially a dependency tree of claims.

This is because the examiner must judge each claim on its own merits, so while the broader claims could get knocked out, the narrower claims could survive and can indicate which "territory" is still unclaimed. The applicant can then focus on that (e.g. by making it part of the root independent claim.)

So often what is interesting in patent applicitations are the dependent claims.

For this patent, the current primary claim literally covers the concept of a tool call by an LLM, and as comments point out, this is ridiculously broad and will get knocked out by prior art. But the dependent claims are intriguing.

A quick glance at those indicates that Mistral thinks the more novel or non-obvious aspects are a "stateless resumable sandbox" which works by executing the generated code until a "non-deterministic operation" (like a random number or a date) and on getting the result of that operation replaying all the code that executed up until that point. The results of previous such operations are cached for reuse in subsequent re-executions.

Now I'm not at all knowledgeable about this area, but I did write a very rudimentary tool-calling LLM harness at one point, and offhand I would now have considered this approach -- re-executing the same code multiple times just seems wasteful and unnecessary. But the specification points out there are other considerations at play in more complex situations, so maybe this is narrower and more interesting than what the title and top claim indicates.

The irony is that it's not probably not patentable in Europe.

The ironic part of it is, that to my own experience and benchmarks Mistral is by far the worst model from frontier labs at tool calling followed closely by Gemini.

Can't beat them... litigate?

I read somewhere that the US Patent Office is like a candy store for other countries to pick up cool tech and continue local development.

Priority date of March 4 2026 and claims this patent is novel on that date, clearly trying it on. Hope it gets struck down.

What are the chances Mistral used AI for its patented technology, thereby potentially invaliding the claim?

Mistal transforms into a patent troll company? Patents organization is broken if it is possible to do

Typically, companies file these patents to protect themselves against patent trolls.

Still pending, or am I mistaken? Can anyone with knowledge on the US process make an estimate how likely this is to go through? I know of Uniloc and their "efforts", is there a possibility that this could take a similarly damaging path?

TONS of prior art.

I don't see anything unique nor novel to what everyone has been doing since GPT-3.5, before OpenAI introduced function calling, I was building my own JSON and tool calling.

Let's find an LLM that was trained before the patent was filed. If it can come up with the idea, then it is trivial and the patent can be discarded.

Mistral has set off my stink test for a while now, they seem to be a "friendly EU AI provider" who are really an EU mouthpiece for regulation and stifling competition instead of actually innovating

Edit: downvote if you want, but here's another example: "Companies selling artificial intelligence models in Europe should pay a "levy" to support cultural industries, the head of French developer Mistral said on Friday, March 20." https://www.lemonde.fr/en/international/article/2026/03/20/m...

what an insane thing to have a patent for.

Cocorico !! Well played Mistral ! Really funny since software patents does not exist in France ...

EU AI innovation is really amazing.

Can somebody care to explain if the 2013 and 2006 refer to IDs or reference to past years.

But... why? Isn't that what smolagents[0] has done from the beginning? Also this move is blackening Mistral in my book.

[0] https://github.com/huggingface/smolagents

If you are a software developer and you sell your time and IP for money to employers. I highly recommend not looking at patents if you don't want to reduce your market price by introducing the liability of patent infringement. Especially don't publicly comment about the patent, there's almost no upside.

The idea of patents is not to read a feed of the new ones and to discuss them as innovations, it's for the patent holders to use as proof in the future that they had already thought about a thing. The best strategy, in any case, is to publish your own patents, or to document your inventions privately or through your own public channels. Such that in the future IF you are sued for patent infringement, you can prove you didn't arrive at your IP by stealing theirs.

Obligatory: I Am Not A Lawyer