MCP Security Starts With Knowing Who Said No
Three layers can refuse an MCP tool call. See why the source field matters for debugging, least privilege, useful audit trails, and AI agent governance.
Read article →Notes, deep dives, and updates from the Bridge GPT team.
Three layers can refuse an MCP tool call. See why the source field matters for debugging, least privilege, useful audit trails, and AI agent governance.
Read article →
A measured guide to inspecting SFCC object attributes with OCAPI, so you can avoid bad assumptions about type, cardinality, localization, and response shape.
Read article →
A public SFCC cartridge shows how four layers of refinement caught contradictions, iterator leaks, and prompt bugs that passing unit tests could not see.
Read article →
See a measured Claude Code and Jira MCP loop from ticket to merged pull request, plus the passing test suite that still missed a live API bug after merge.
Read article →
A wrong practice question is worse than none. Here is how we built a machine that assumes its own exam questions are broken until a second, hostile pass proves otherwise.
Read article →
A 60-question Salesforce Admin practice exam only helps if the 60 match the real topic weighting. Here is the current blueprint and the exact allocation it implies.
Read article →
A historically grounded look at whether AI will destroy knowledge work, drawing on the ATM paradox, the handloom weavers, and the honest economics of displacement.
Read article →
Navigate the 2026 MuleSoft certification path. Learn the smart order to take exams, which retiring certs to avoid, and where to find free practice tests.
Read article →
A practitioner's honest breakdown of the best free and paid study resources for the 2026 Salesforce Platform Administrator exam.
Read article →