Is this basically a LLM that has tools automatically configured so I don’t have to handle that myself? Or am I not understanding it correctly? As in do I just make standard requests , but the LLM does more work than normal before sending me a response? Or I get the response to every step?
spmurrayzzz · 15h ago
The aspirational goal is that the model knows what tools to call and when, without human intervention. In practice, you'll see varying efficacy with that depending on the tools you need. Some of the tool usage is in-distribution / well represented in training set, but if you have some custom exotic MCP server you created yourself (or pulled off of some random github) you may see mixed results. Sometimes that can be fixed by simply augmenting your prompt with contrastive examples of how to use or not use the tool.
As an aside, my experience with devstral (both via API and locally w/ open weights) has been very underwhelming to this effect. So I'm curious how this new agent infra performs given that observation.
koakuma-chan · 15h ago
It's a software framework for orchestrating agents. Each agent can have its own system prompt, its own tools, and it can delegate ("hand off") to a different agent. When a hand off occurs, the LLM runs again but as a different agent.
manmal · 14h ago
Like Gemini Gems, but agentic?
koakuma-chan · 14h ago
Gemini Gems seems to be a ChatGPT “GPTs” equivalent, and I never figured out what those actually are. Mistral Agents API is like OpenAI Agents SDK.
LeoPanthera · 13h ago
Gems and GPTs are just a way to customize the system prompt from the web UI.
qwertox · 13h ago
The "My MCPs" button looks very promising.
I was looking around at Le Chat, a thing I haven't done in months, and I thought that they've really worked on interesting stuff in interesting ways.
The ability to enrich either a chat or generally an agent with one or more libraries has been solved in a very friendly way. I don't think OpenAI nor Anthropic have solved it so well.
bbor · 16h ago
Ok I’m behind the times in terms of MCP implementation, so would appreciate a check: the appeal of this feature is that you can pass off the “when to call which MCP endpoint and with what” logic to Mistral, rather than implementing it yourself? If so I’m not sure I completely understand why I’d want a model-specific, remote solution for this rather than a single local library, since theoretically this logic should be the same for any given LLM/MCP toolset pairing. Just simpler?
It seems like the main pitch here is auto-inclusion and auto-exclusion of various tools via an orchestration agent (which may or may not be the main model itself? Unclear from their post)
Mostly this seems like an end-run around tool calling scalability limits. Model performance degrades heavily if the field of possible tools gets too large, so you insert a component into the system that figures out what tools should be in-scope, and make only those available, to get reliability higher.
In terms of "why outsource this" it seems like the idea is that their orchestration agent would be better than a cruder task state machine that you would implement yourself. Time will tell if this assertion is true!
ed · 11h ago
> auto-inclusion and auto-exclusion of various tools via an orchestration agent
Where do you see that? That would be neat, I'm under the impression orchestration is manual though – you define an agent and give it the ability to hand off tasks to sub-agents.
orliesaurus · 16h ago
Whoever made those embedded videos, here some feedback if you want it take it, it's free:
1) It's really hard to follow some of the videos since you're just copy pasting the prompts fr your agents into the chat because the output generation comes out and hides the prompts. Instead put the prompt text as an overlay/subtitle-like so we know what you're doing
2) The clicking sound of you copy pasting and typing is not ASMR, please just mute it next time
3) Please zoom into the text more, not everyone has 20/20 super vision 4K style
threeducks · 16h ago
To add to 3): YouTube embedded videos default to 360p for me even if I maximize the embedded video on my 4k screen, which is completely unreadable. This is probably an attempt by YouTube to get viewers to click through to the YouTube website. It is probably not in Mistral's best interest to funnel viewers to YouTube, so they should use a different video host.
But even at maximum 1080p resolution, the image quality is not that great. And while we are at it, the wine-red (#833048) on dark-brown (#23231F) syntax highlighting for keyword arguments has very poor contrast ratio of around 1.8 to 1: https://webaim.org/resources/contrastchecker/ which earns a rating of "Fail" across the categories normal text, large text and UI elements.
ianhawes · 16h ago
4) Use a clean browser profile so you don't show unrelated autocomplete
moralestapia · 13h ago
I came here to see if anyone else noticed.
Very sloppy job, imo.
It costs next to nothing to come up with a little story and have someone on Fiverr narrate it (or an AI, after all that's what they sell).
htrp · 16h ago
is mistral a model company, an agent company, or a enterprise software company now?
brandall10 · 14h ago
Couldn't the same questions be asked of OpenAI and Anthropic?
Ultimately these are product/service companies, levering their research and innovations as differentiators.
If you're "only a model" company you likely have no moat.
greenavocado · 15h ago
Mistral is trying to be everything at once and it shows. To make ends meet they pivoted to selling enterprise software through Le Chat and cozying up to Microsoft. Now they're throwing around terms like "agentic AI" to stay trendy, even as competitors like DeepSeek outperform them in key areas. Their identity crisis is obvious. Are they a model company? A software vendor? A research lab? At this point, they seem more like a startup chasing hype and funding than a company with a clear direction. The 6 billion Euro valuation looks impressive, but with so many shifts in strategy, you have to wonder if they're building something lasting or just riding the AI wave until it crashes.
eigenspace · 10h ago
Their strategy doesn't make sense to you because you're looking for a technical feature that differentiates them. But technical features aren't their key differentiator, geography is their key differentiator. They'll get a lot of contracts in Europe simply because they're European. Everyone is keenly aware of how dependant European tech stacks are on increasingly unfriendly foreign powers.
If there's a local European option that does most of what an American or Chinese company does, that's simply a safer choice.
From this point of view, them trying to do everything at once makes a lot of sense. They don't actually need to be the absolute best or even the cheapest at any one thing. They need to just exist in Europe, be stable, and offer good services that people want. Casting a wide net is a better strategy for them.
Raed667 · 15h ago
Do they need to pick one? Their offering doesn't seem incoherent to me
As an aside, my experience with devstral (both via API and locally w/ open weights) has been very underwhelming to this effect. So I'm curious how this new agent infra performs given that observation.
I was looking around at Le Chat, a thing I haven't done in months, and I thought that they've really worked on interesting stuff in interesting ways.
The ability to enrich either a chat or generally an agent with one or more libraries has been solved in a very friendly way. I don't think OpenAI nor Anthropic have solved it so well.
It certainly looks easy to implement, I will say that! Docs halfway down the page: https://docs.mistral.ai/agents/mcp/
Mostly this seems like an end-run around tool calling scalability limits. Model performance degrades heavily if the field of possible tools gets too large, so you insert a component into the system that figures out what tools should be in-scope, and make only those available, to get reliability higher.
In terms of "why outsource this" it seems like the idea is that their orchestration agent would be better than a cruder task state machine that you would implement yourself. Time will tell if this assertion is true!
Where do you see that? That would be neat, I'm under the impression orchestration is manual though – you define an agent and give it the ability to hand off tasks to sub-agents.
1) It's really hard to follow some of the videos since you're just copy pasting the prompts fr your agents into the chat because the output generation comes out and hides the prompts. Instead put the prompt text as an overlay/subtitle-like so we know what you're doing
2) The clicking sound of you copy pasting and typing is not ASMR, please just mute it next time
3) Please zoom into the text more, not everyone has 20/20 super vision 4K style
But even at maximum 1080p resolution, the image quality is not that great. And while we are at it, the wine-red (#833048) on dark-brown (#23231F) syntax highlighting for keyword arguments has very poor contrast ratio of around 1.8 to 1: https://webaim.org/resources/contrastchecker/ which earns a rating of "Fail" across the categories normal text, large text and UI elements.
Very sloppy job, imo.
It costs next to nothing to come up with a little story and have someone on Fiverr narrate it (or an AI, after all that's what they sell).
Ultimately these are product/service companies, levering their research and innovations as differentiators.
If you're "only a model" company you likely have no moat.
If there's a local European option that does most of what an American or Chinese company does, that's simply a safer choice.
From this point of view, them trying to do everything at once makes a lot of sense. They don't actually need to be the absolute best or even the cheapest at any one thing. They need to just exist in Europe, be stable, and offer good services that people want. Casting a wide net is a better strategy for them.