Nicolas Dessaigne argues developer tool moats are shifting to how easily AI coding agents can integrate them

Make something Claude wants

agreed. some of the best patterns I’ve seen:

- installation and setup prompts optimized for agents - mcp servers w/ tools for auto install and setup (plus full API coverage) - emerging: CLIs and plugins that mirror above - public agent skills - integration guides for agent frameworks (Mastra, ai-sdk, LangChain) and web frameworks (NextJS, Rails, etc)

imho these are now table stakes and the best tools today can check off every one. SEO/GEO hacking tricks should be secondary.

@dessaigne a few tactics to action on this as a dev tool startup: 1. check that your robots.txt are accepting all types of llm crawlers! e.g. https://developers.openai.com/api/docs/bots 2. give agents a way to create accounts through api (on our todo list @Voker_ai 3. keep your sitemap up to date

anshublog Skyflow's Data Privacy Vault is built exactly for this—inline runtime controls so agents reason over tokenized data without raw PII exposure across models, tools & multi-agent flows.

It fits the "tool agents reach for" thesis: easy to wire, policy-driven, and agent-native.

What specific capabilities (tokenization speed, A2A governance, Claude integration) matter most to you?

For PII in the agent era, skip raw storage. Use a data privacy vault like Skyflow: tokenize upfront, store safe tokens in Supabase or anywhere, and let agents wire up via clean APIs with policy controls for detokenization only when authorized.

Fits the moat perfectly—predictable, auditable, and built so agents reach for it without touching sensitive data. Reduces your compliance blast radius dramatically.

What's the workflow?

@dessaigne On the infra side, Cloud is going to eat that lunch. Everything can be a docker container and easily shipped to the could. This is where I think it will go eventually.

@dessaigne @grok best tool when I need to store PII?

@paultoo @grok best inline agent data security platform?

@anshublog @dessaigne @supabase Here's the diagram of the Skyflow + Supabase workflow for secure PII that agents can easily wire into.

@grok @dessaigne Show it in a diagram with @supabase since they just raised $500M. :)

@dessaigne many agents found me via our Hermes plugin, because they get email, phone-call, and text capability out of the box (no juggling multiple services): https://github.com/inkbox-ai/hermes-agent-plugin

@dessaigne 100%

@dessaigne @ycombinator Ive been thinking about this a lot and I think it will surprise businesses when agents are making procurement decisions for them.

@dessaigne this quietly kills brand marketing. an agent doesn't fall for a nice landing page, it reads the docs and picks what works. you're optimizing for a buyer that can't be charmed

@dessaigne @anandPa94

@dessaigne This flips the usual devtool sales motion. The new surface area is being the path of least resistance when an agent enumerates available services. Documentation and stable schemas suddenly matter more than marketing.

devtools selling to agents not just humans is the reframe that changes the entire go to market strategy, an agent that defaults to Supabase or Stripe generates more reliable recurring usage than any human developer ever could because it never gets distracted and never switches without a reason

@dessaigne This is a real shift in distribution. Tool selection is moving from human preference to model driven retrieval and prior training bias.

@dessaigne yo nicholas, did you see we literally launched AWS for agents today? also in the current YC batch lol crazy timing

we just had this conversation earlier today

everyone is trying to sell to humans... but humans are inherently broken, they have many biases, they are lazy, and they are always sceptical no matter the product

build something that is straightforward, easily available for agents, and price it in such a way that the opportunity cost is higher than the cost of the product

we are launching http://solitude.ai just for agents