MuleSoft's undocumented connectors are a liability in the LLM era
MuleSoft’s poorly documented connectors — GraphQL, Snowflake, Salesforce (in my personal order of annoyance) — don’t just land in the now overflowing “Salesforce doesn’t GAF about customers or partners” bucket 🪣.
They also make MuleSoft a liability in an LLM-assisted dev world.
Mule apps should be perfect for LLM-assisted coding: consistent (albeit XML 🤮) structure, types and optionally schemas for all inputs/outputs, and connectors that represent well-documented protocols and coding concepts.
But MuleSoft couldn’t be bothered to fully document their Official Connectors, despite the price tag of a timeshare in Lake Tahoe. The basics tend to be covered, but as you get more into the weeds, it's just sad empty boxes where descriptions and syntax should be.
The ironic bit? MuleSoft provides an MCP connector so customers can generate Machine-readable Component Plans (MCPs) to help AI tools assist with building integrations… but doesn’t offer an MCP to help developers build MuleSoft apps. 🫠
Where’s that “AgentForce! AgentForce! AgentForce!” AI energy, MuleSoft?
And it's not just Mulesoft, I’ll believe AI has become sentient and taken over Salesforce's development when:
😤 They get past the weird limitations of their most ancient standard objects
⚡ Lightning pages are more... lightning
🤭 12 year old "popular" Ideas get implemented
Until then, “AI-forward” is just the marketing team talking to itself.
#Salesforce #MuleSoft #AI #DevExperience #LLM #Codality
Jessie Grenfell
Jessie has spent 20+ years architecting and delivering software for organizations where complexity is the norm: financial services, insurance, nonprofits, and regulated industries where getting it wrong isn’t an option. She specializes in the hard problems: unique operational workflows, cross-system integrations, and compliance requirements (HIPAA, CCPA, SOC 2) built into the design from the first conversation.
Her AI work comes out of that background. She has shipped AI to production, not just piloted it: a multi-agent research pipeline that cut a client’s investigation time from over an hour to minutes, and natural-language analytics that let executives ask plain-English questions of their operational data and get validated charts back. Years building business critical systems taught her where AI belongs in a business process and where it doesn’t.
At Codality, she leads engagements from discovery through delivery, because she’s seen what happens when that thread gets cut.