Launching today
Navige
Cut AI agent costs and control runaway spending
48 followers
Cut AI agent costs and control runaway spending
48 followers
Navige uses Sufficiency Intelligence to cut AI agent costs while work runs. It sends suitable requests to smaller models, caches repeated reads and trims bulky tool results. If unsure, it keeps your chosen model. Set budgets, catch runaway loops and approve sensitive actions. See costs and savings by job, with estimates clearly labelled. Watchtower is free forever. Preview changes before switching controls on.
Hey Product Hunt 👋
Why we built Navige
Most of what an AI agent does isn’t hard - sorting an email, pulling a date out of a document, picking the next step. But often, it all runs on the most expensive model, because that’s the one written into the code and nobody went back to change it.
Then the bill arrives as one number, and nobody can say which agent spent what.
Navige sits in front of calls routed through it and acts before they run, while the cost can still change. We call this Sufficiency Intelligence: give each task exactly the intelligence it needs.
How it works
It assesses each request and sends suitable ones to a smaller model. Your model is the ceiling: Navige only moves calls down, never up, and keeps yours whenever it’s unsure.
If the smaller model fumbles a tool call, the call re-runs on your original model, and your agent only sees that answer.
It records every job: what it cost, what the controls changed and what made the difference. Measured savings and estimates are clearly separated.
On top of that, you get budgets per team and per job, approval holds for risky actions, a kill switch and an audit trail.
Two things we did on purpose
Watchtower is the starting point. Navige watches and shows what it would have done, but changes nothing until you say so.
Connect your Anthropic or AWS Bedrock account and put your real bill next to Navige’s figures, so you can check the numbers instead of trusting a savings claim.
Pricing
Watchtower is free forever. On paid plans, of the measured savings Navige can show, you keep four parts and we keep one, subject to a $9/month minimum.
Give it a week
Watch your agents for a week. Nothing changes until you decide.
Which of your agents’ jobs do you suspect is running on a much bigger model than it needs? 👇
@sojimathewj The project sounds amazing; Quick question on my end: What if my agent asks the same question twice? Does Navige charge for both calls?
@dounia_boughalem Not necessarily! If Navige can safely reuse the earlier answer, there is no need to call the model again and pay twice for the same work.
I like the idea! But what happens if Navige chooses a cheaper model and the answer is worse? That's a risk I see here
@vlad_popov3 That is a fair point, and something we have thought a lot about. The idea is not to just switch to a cheaper model to save money. If a task needs a powerful model, then it should absolutely use one. But not every task needs that level of intelligence. The tricky part is knowing where to draw that line without affecting the quality. That is exactly what we are trying to solve with Navige. No point saving a few dollars if you end up with a rubbish answer!
Hi all, Akhil here, co-founder and CCO. Soji’s covered how Navige works, so I’ll talk about the money side.
If you sell an AI product on a flat monthly price, you don’t really control what each customer costs you. Someone paying $30 a month can burn through more than that in model calls, and you won’t know until the invoice comes in. A lot of that spend is simple stuff, all running on a model that’s far bigger than the job.
Navige shows you what each agent and each job is actually costing you, and where a smaller model would have done fine. You can also set a budget, so a loop that goes wrong gets stopped before it shows up on your bill.
We start everyone on Watchtower for that reason. You see the real numbers first and nothing changes until you decide.
Genuinely curious...... if you are running agents in production, do you know what a single job costs you? Or is it a surprise at the end of the month?
@akhil_shajahan Well said, Akhil! Nothing hurts more than finding out your customer spent more on AI than they actually paid you. That is one surprise we would love to help businesses avoid!
What's the catch with Watchtower being free forever?
@vishal_sambath Pretty much as soon as your agent starts making calls through Navige. You can see what it is doing, which models it is using and how much those calls are costing you. The more your agent runs, the more useful the data becomes.
We also send a Watchtower report every Monday which would highlight areas where you might be spending more than necessary. So you don't need to do much beyond connecting your agent to get started.
Hellooo Product Hunt!!
I'm Shameeh; I handle solutions and customer success at Navige.
@sojimathewj Covered the "why". I'll cover the "what happens next" since that's where I come into play.
The first week is Watchtower (which is free forever). You connect all your agents' calls through Navige as usual. The way everything works for them stays the same. Then you continue using your agents as usual, but this time, you'll see exactly how much each individual agent paid, and what Navige would have done differently.
Every week, you will receive an email with details. Once you are ready, switch to the paid plan, and actual savings start to pour in.
And one other thing: your agents can be one or a team; we can handle it.
Now, if your business has some specifics (a number of different agent frameworks, shared keys between teams, an invoice no one can understand), I would love to hear about it. Leave a comment below or ping me directly; I'll get you connected.
And if you try the service and there's something confusing about it, do tell. That's the best feedback we can get today.
If I don't reply (which I highly doubt it), there are few others looking at their screens to help you guys out, please send them a text and put them to work as well. @sojimathewj @akhil_shajahan @alanantony
@akhil_shajahan @alanantony @shameeh Thanks Shameeh! I think we have just found ourselves a volunteer to handle all the tricky customer questions today! Jokes aside, love the way you have explained it. Over to you!
Excited to be hunting Navige today!
What caught my attention is how simple the idea is: most of what AI agents do doesn't need the most expensive model, but that's usually what runs every single call. Navige gives each task just the model it needs, only ever moves calls down, and falls back to your original model if a smaller one fumbles. Plus budgets, approval holds and a kill switch for when an agent goes off the rails.
What I love most is Watchtower, which is free forever and just shows you what it would save before you switch anything on. Congrats @sojimathewj and the team on the launch! Let's go!
@kate_ramakaieva Thank you so much for hunting Navige today! You have probably explained what we do better than I have in the last 13 months! 😂 Really appreciate your support and for being part of our big day. Means a lot to all four of us. Let’s go!
After many months of building, breaking things, fixing them, and occasionally wondering why something worked yesterday but not today, Navige is finally live on Product Hunt! As the person leading engineering, it is pretty exciting to see all those late nights turn into a real product. Huge credit to @sojimathewj , @akhil_shajahan and @shameeh for this crazy journey together.
We have built something we genuinely believe can change how people use and pay for AI. Now comes the fun part, seeing what the world thinks!
Give Navige a spin. Would love your feedback, especially from fellow builders!