CLIs & Agent Skills are becoming the new interface of SaaS. 

Not just for humans, for agents that discover, evaluate, and operate products on their behalf.

Resend, an email solution for developers, launched in 2023 with zero users. Today it has 500,000 developers. 

SendGrid has been around since 2009, Twilio paid $3B for it, and it has 15 years of track record. 

When Claude Code picks an email provider, it chooses Resend 63% of the time. SendGrid, only 7%.

Something different is happening in SaaS.

A new 'user' is here

For two decades, we've designed software for humans clicking through interfaces. Now there's a new 'user' in this AI-native world: agents (LLMs with access to tools) that discover, evaluate, and start using products without ever touching your UI.

🔹 They don't read your landing page, they read your API 

🔹 They don't try your trial, they try your CLI 

🔹 They aren't convinced by your sales motion, they're convinced by your documentation

As Brendan Short put it: «Software discovery now kicks off from Claude. This is a different buying motion than anything we've seen in B2B SaaS».

What this means for SaaS

The agent doesn't care that you've been around for 15 years. It cares whether your API is clean and your docs are readable. 

From SEO, SDRs, and free trials… to the agent's judgment. CLIs, MCP, and Agents are emerging as a new distribution channel, and the window to explore it is still wide open.

Agentic UX: Designing for agents

For the past two decades, software design has been almost entirely a conversation about humans, information architecture, UX research, design systems, accessibility. Every discipline assumed a person clicking, scrolling, typing.

That assumption is no longer the only one worth designing for.

We've spent years polishing UX for people. Now it's time to build the equivalent layer for agents. Call it Agentic UX:

🔹 Well-documented APIs, designed so an LLM can understand them in one pass 

🔹 CLIs that expose the full functionality of the product, not just a subset 

🔹 Agent Skills and MCP servers that make your systems interoperable

Marc Benioff said it clearly: «Our API is the UI». Salesforce has already exposed its entire platform, apps, workflows, metadata, Agentforce, Slack, as unified APIs, MCP tools, and CLI.

CLIs, MCP, and Skills: the bridge between systems

An API alone doesn't make your product agentic. What truly unlocks value is the combination: API + CLI + Agent Skill + MCP. That layer lets the agent close the full loop within a single workflow:

🔹 Read data → query your system in real time, no manual exports 

🔹 Analyze → apply LLM models to extract insights or run specific routines 

🔹 Act → write back into your product: create records, trigger tasks, update statuses

The agent stops being an assistant and becomes an operator that executes tasks across systems. This isn't about "adding AI to the interface", it's about your product being orchestrated alongside any other tool the agent has access to.

The growth window for incumbents

Established SaaS players have the context and the data. What they're missing is the agentic layer. Whoever builds it first wins the agent's workflow within their category. 

Agentic UX doesn't replace human UX, it expands who we design experiences for.

That shift just reached the coworking stack.

Nexudus has made its platform fully accessible to agents through a stronger API, a purpose-built CLI, and a structured skill layer that teaches agents how to use both. It changes how coworking operators can use AI.

«Anything you can do in Nexudus can now be orchestrated programmatically by an agent and enriched by any other tools and data the agent can access.» — Adrian Palacios

In the latest edition of Coworkings AI, we break down what the Nexudus CLI & Agent Skill are, why they matter, and walk through a hands-on guide: turning member feedback into a prioritised action plan, end to end.

The workflow: 

🔹 Claude pulls live helpdesk data from Nexudus 

🔹 Runs a full Member Feedback Analysis Claude Skill — topic extraction, sentiment analysis, priority matrix, 30/60/90 plan 

🔹 Pushes tasks back into Nexudus for the ops team to review and act on

A closed loop, end to end, from a single agent routine. And member feedback is just the start, the same pattern unlocks inbound sales and lead nurturing, member onboarding, churn signals, billing issues, adoption, and cross-selling.

Agents handle the repetitive ops. Operators get back to the members.

Full article here: 👇 https://coworkings.ai/p/coworkings-ai-011