Propane gives your product team and agents one connected, always-current view of your customers. Automatically collected from all your tools. Collaborate on a shared canvas. Commit straight to any coding or design agent. Secure, maintained, always on. You just build products people love.
I've been watching what you've been building for a while now and is genuinely excited for this.
With the teams you've been testing with, where do teams typically see the biggest lift first: support speed, product discovery, or better handoffs between teams?
And how do you see your tool stack up against, for instance, harvestr?
Hej @toxboe 👋 Thanks for following along!
I'd say the biggest lift is actually increase alignment, because they have a shared context and understanding from the get go.
Discovery comes second, and something that sneaks up on the teams, as they get more and more invested in the platform.
I haven't used Harvestr, but did a quick background check, and I'd say it's a different tool, for a different problem. They look like they are more focused downstream, with collecting and organizing feedback, and pushing that to a roadmap, right? We're focusing more upstream, and signals come before feedback, and the context is there for the agents, when you want to dig in. So we overlap, but we've taken a different approach to it.
Happy to show you what we're seeing.
Great work and congrats on the launch. Am I correct to assume the always-current, auto-collected part is the hardest element? Wondering what happens when two tools disagree about the same customer (ex. - sales notes say one thing, support tickets another) - what do you treat as truth before committing that to an agent?
@artstavenka1 Yes, you're right, that's the hardest part. Think of it like a data warehouse problem: deduplication, matching, conflict resolution across sources. That's real infrastructure work and most teams have no business spending cycles on it.
Our view is that teams shouldn't have to deal with that layer at all. You connect your tools, you get access to clean, enriched context. We handle the hard job underneath. The goal is that when two sources disagree, that's our problem to resolve, not yours.
automatic customer context is the thing every product team fakes with scattered notion docs. does it pull from support + sales + product usage, or mostly one source?
@reallynattu hey there nattu thanks for sharing!
agree, so many people are trying build & maintain their own system in notion, but it ends up owning them instead with the amount of maintenance!
yep support, sales (crm) and product - intercom, hubspot, posthog as some examples:)
we pull stuff from these sources, filter the signal from the noise so you have simple product context you can work with across your team
@reallynattu No we support a range of sources and will expand on that, but CRM, Support, calls etc we do full indexing for now, then we add layers with Product usage via MCP but plan to index that all in to one context.
What tools are you looking for to index? what your stake like?
Introducing one flat price for unlimited users and tools is a huge win for lean startups that are sick of seat-licensing fatigue, Congrats 👏 for shipping. qq does the base plan track historical context storage limits, or are we literally only billed on the net-new data ingested during the billing cycle?
@vikramp7470 Good question. Historical context is included in the base plan, no extra charge. You only get billed when we index a new customer or competitor for the first time in a billing cycle. After that, every update, signal, or data point we pull on that same customer counts as one index event, not one per update. So 30 signals on a customer in a month still equals one billing unit for that customer that month.
@greenlieber Thanks for clarifying, super fair pricing model, appreciate the breakdown.
When a customer's context changes after a coding agent has already committed work based on the older context, does Propane flag that drift, or is keeping the agent's snapshot up to date on the team?
@nikita_zaro thanks for your comment nikita!
right now propane has a two-speed system of ingesting context:
signals - which is continuous, proactive and surfaces what matters
mcp - where the agent can retrieve what it needs on-demand (up to date, latest)
so right now, it kinda depends which tools you have connected!
you would definitely see what customers are saying about whatever you just committed/shipped coming in via signals with your crm, intercom etc. hooked up. however, we are working on ways of getting a bidirectional context so propane has an explicit awareness of what you shipped from our platform
our vision is to make product context simple and effortless for the team:)
@benfleming Thanks for the thorough breakdown! The signals + MCP split makes sense, and bidirectional context awareness sounds like a powerful direction. Excited to see where you take it.
@nikita_zaro anytime:) yeah we are working on it!!
let me know if you are interested to got a bit deeper, and we can take you through the product with a founder @greenlieber I'm sure would!
@greenlieber very cool, congrats on the launch!! Would you mind going a bit deeper into to the type of information / data that is fed into the product? You touched on it in the video but would love to hear a little bit more about that. Thanks!
@millwiller So we can pull from crms, calls, support etc. focus right now is qualitative data that we index canonical, but we have a plan to add quantitative to the profiles of users and customers, So you can go deeper.
So context is not only a layer one side but also with more rich attribution; t
What type of data are you looking to have? from a tool or data you dont have today?
@greenlieber That’s super helpful! Think the direction you’re going in makes a lot of sense - starting with qualitative (more difficult probably) then adding in quantitative later. We are extremely early stage B2C so don’t have a robust CRM solution in place but something like this seems super helpful with targeted marketing.
@millwiller intersting and true that what we build had some level of an CRM
Congrats on the #1 spot today! Automatic context for agents is such an underrated problem — most teams are still manually re-explaining the same context every time. Curious what was the trickiest part of keeping that context accurate as products evolve?
@martin_mo appreciate that Martin!!🤝
100% this is a problem we have been grappling with ourselves, and for sure really useful product context is buried in the tools that users already use, but often scattered and inaccessible!
One tricky thing in the AI-era there is dealing with a lot of noise - it's not just finding the important product context, but also working out what actually matters and what is emerging across customers..
tackling this was hard, but we think we have gone to great efforts to make this simple with signals. we constantly dig through your customer feedback, product analytics etc. to give you the most important product context at your fingertips!
what have you found challenging as a builder building SURO AI in this era!?
@benfleming "Really appreciate the detailed take — and the signal-vs-noise framing resonates, though as a solo builder I'd say I'm dealing with the inverse problem: not too much data, but too little. With a small early user base, it's genuinely hard to tell whether something is a real pattern or just noise from a handful of people. I've had to lean more on direct conversations than analytics at this stage — though I can already see how a tool like yours becomes essential once that scales up. Biggest challenge for me right now is probably wearing every hat at once — product, support, and growth — with no team to delegate any of it to. How did your team decide which signals to prioritize surfacing first when you were validating the idea?"
@martin_mo anytime man!
for sure there are different problems at different stages, however something useful for you may be the competitors tracking, so you can see what others are building and keep on top of that!
sometimes it is hard with so much information out there coming at a rapid pace!
@benfleming That's a great point — competitor tracking is something I've been doing pretty manually so far, mostly just checking in on a handful of tools every week or two. Do you have a lightweight way of doing this yourself, or is it more about setting up alerts for specific moves (pricing changes, new features, etc.) and reviewing those periodically? Trying to find something sustainable that doesn't eat up too much time as a one-person team.