Useful AI devtools can still be fragile companies.
The fastest way to tell the difference is to ask what survives the platform roadmap.
Not "is the product good?"
Not "did developers upvote it?"
Not "does the demo feel sharp?"
Those matter. But they are not the first structural check.
The first check is this:
If the platform underneath this product changes, what customer-owned value remains?
The category is moving too fast for surface reads
Agent devtools now sit on top of layers that change every week:
- coding agents- IDEs- model APIs- MCP servers- Slack and GitHub workflows- browser automation- local runtimes- team-chat surfacesThat speed creates real opportunity. It also creates a very specific diligence gap. A product can be early and useful while still being exposed to a platform owner with more distribution, more data, and a stronger default position. The investor mistake is treating usefulness as durability. The founder mistake is waiting until the meeting to explain the difference. ## The platform-risk read When I screen an AI devtool, I would map the company across four questions. ### 1. What is the platform likely to absorb? Some features are obvious roadmap items. Session visibility for coding agents. Team-level agent logs. Native memory. Connector setup. Cost dashboards. Permission prompts. Browser access checks. A startup can still win there, but the answer has to move beyond polish. The stronger case is usually one of these:
- the product is neutral across platforms- the product owns a workflow across multiple tools- the product serves a buyer the platform does not prioritize- the product has a trust or deployment posture the platform cannot easily copy- the product has accumulated context or usage history that compoundsIf the only answer is "the incumbent will not build it," the case is not ready. ### 2. Which part of the value is customer-owned? This is the cleanest way to separate feature from company. Customer-owned value might be:
- retained team context- audit history- production usage data- deployment-specific policies- internal workflow memory- buyer-specific integrations- security postureA UI around another vendor's workflow is easier to absorb. A system that becomes the customer's operating record is harder to replace. ### 3. What breaks if a connector changes? Connector-heavy tools need a clear data-flow story. For products touching Slack, Gmail, GitHub, docs, or internal communications, the important question is not only "can you connect?" It is: What do you store, process, index, and retain — and under what permission model? A serious buyer will ask it. A serious investor will ask it. A strong founder should have the answer ready. ### 4. What proof exists outside the demo? Launch momentum helps. Demos help. Product copy helps. But platform-risk claims get stronger when the public surface shows independent proof:
- paid teams- retained usage- customer references- usage after the launch spike- evidence of replacing a workflow- public technical docs that explain trust boundaries- clear pricing that reflects the value capturedThat is the difference between "this is useful" and "this may be durable." ## Why this helps both sides The platform-risk question is not hostile. It is a shortcut to a better meeting. For investors, it prevents a common mistake: mistaking a beautiful feature for a company with leverage. For founders, it gives the cleanest preparation prompt: Show what remains yours after the platform notices. That answer can be technical. It can be commercial. It can be workflow-based. It can be trust-based. But it has to exist. ## How CyberFruit turns this into a report CyberFruit looks at the public surface of a startup and turns it into a sharper pre-call read. For agent devtools, that means mapping:
- the product claim- the platform dependency- the data-flow questions- the public proof- the strongest first question to take into the meetingThe output is not a takedown. It is the work both sides should want done before the conversation starts. The best founders are not afraid of this question. They use it to make the company clearer. See how this looks in a live agent-devtool screen -> cyberfruit.ai/curated-reports/2026-06-25-linzumi
Top comments (0)